Using R and OpenBUGS for Evaluating the Causal Effect of Dynamic Treatment Regimes

Size: px
Start display at page:

Download "Using R and OpenBUGS for Evaluating the Causal Effect of Dynamic Treatment Regimes"

Transcription

1 Using R and OpenBUGS for Evaluating the Causal Effect of Dynamic Treatment Regimes Daniel Scharfstein and Chenguang Wang June 9, 2010 C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 1/42

2 Outline 1 Acute Lung Injury and ARMA Trial 2 Bayesian Causal Inference 3 R and BUGS 4 Application Results of the ARMA Trial 5 Concluding Remarks C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 2/42

3 Acute Lung Injury (ALI) C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 3/42

4 Acute Lung Injury (ALI) ALI is a condition characterized by acute onset of Shortness of breath Rapid breathing rate Low oxygen levels Abnormal chest x-ray C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 3/42

5 Acute Lung Injury (ALI) ALI is a condition characterized by acute onset of Shortness of breath Rapid breathing rate Low oxygen levels Abnormal chest x-ray Adult respiratory distress syndrome (ARDS) is a subset of ALI with more severe hypoxemia. C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 3/42

6 Acute Lung Injury (ALI) ALI is a condition characterized by acute onset of Shortness of breath Rapid breathing rate Low oxygen levels Abnormal chest x-ray Adult respiratory distress syndrome (ARDS) is a subset of ALI with more severe hypoxemia. ALI can be a direct injury to the lung (e.g. pneumonia) or indirect lung injury (e.g. pancreatitis). C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 3/42

7 Acute Lung Injury (ALI) ALI is a condition characterized by acute onset of Shortness of breath Rapid breathing rate Low oxygen levels Abnormal chest x-ray Adult respiratory distress syndrome (ARDS) is a subset of ALI with more severe hypoxemia. ALI can be a direct injury to the lung (e.g. pneumonia) or indirect lung injury (e.g. pancreatitis). Several (meta-analysis) studies have found that mortality rate for ALI patients is about 40%. C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 3/42

8 Acute Lung Injury (ALI) ALI is a condition characterized by acute onset of Shortness of breath Rapid breathing rate Low oxygen levels Abnormal chest x-ray Adult respiratory distress syndrome (ARDS) is a subset of ALI with more severe hypoxemia. ALI can be a direct injury to the lung (e.g. pneumonia) or indirect lung injury (e.g. pancreatitis). Several (meta-analysis) studies have found that mortality rate for ALI patients is about 40%. C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 3/42

9 Mechanical Ventilation for ALI C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 4/42

10 Mechanical Ventilation for ALI ALI patients often require mechanical ventilation in an intensive care unit (ICU) to help them breathe. C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 4/42

11 Mechanical Ventilation for ALI ALI patients often require mechanical ventilation in an intensive care unit (ICU) to help them breathe. C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 4/42

12 Mechanical Ventilation - Key Terminology C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 5/42

13 Mechanical Ventilation - Key Terminology C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 5/42

14 Mechanical Ventilation - Key Terminology Positive end-expiratory pressure (PEEP): the alveolar pressure at the beginning of the breath C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 5/42

15 Mechanical Ventilation - Key Terminology Positive end-expiratory pressure (PEEP): the alveolar pressure at the beginning of the breath Plateau pressure (PP): pressure experienced by lung at end of each breath C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 5/42

16 Mechanical Ventilation - Key Terminology Positive end-expiratory pressure (PEEP): the alveolar pressure at the beginning of the breath Plateau pressure (PP): pressure experienced by lung at end of each breath Tidal volume (VT): Size of breath given to patient from ventilator (set by doctor) C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 5/42

17 Mechanical Ventilation - Key Terminology Positive end-expiratory pressure (PEEP): the alveolar pressure at the beginning of the breath Plateau pressure (PP): pressure experienced by lung at end of each breath Tidal volume (VT): Size of breath given to patient from ventilator (set by doctor) Lung compliance (LC): stiffness of lungs (increases with severity of ALI) VT LC = PP PEEP C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 5/42

18 ARMA Lower Tidal Volume Trial C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 6/42

19 ARMA Lower Tidal Volume Trial The ARMA study was a randomized, controlled multi-center 2 x 2 factorial study conducted by the National Heart, Blood and Lung Institute (NHBLI) ARDS Clinical Network C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 6/42

20 ARMA Lower Tidal Volume Trial The ARMA study was a randomized, controlled multi-center 2 x 2 factorial study conducted by the National Heart, Blood and Lung Institute (NHBLI) ARDS Clinical Network The study consisted of a drug treatment (Ketoconazole vs. placebo) and a ventilation strategy (6ml/kg tidal volume vs. 12ml/kg tidal volume) C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 6/42

21 ARMA Lower Tidal Volume Trial The ARMA study was a randomized, controlled multi-center 2 x 2 factorial study conducted by the National Heart, Blood and Lung Institute (NHBLI) ARDS Clinical Network The study consisted of a drug treatment (Ketoconazole vs. placebo) and a ventilation strategy (6ml/kg tidal volume vs. 12ml/kg tidal volume) A total of 432 patients were randomized to lower tidal volume ventilation arm. C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 6/42

22 Lower Tidal Volume Ventilation (LTVV) Set initial tidal volume to 6 ml/kg predicted body weight (PBW) Decrease VT to 5 or 4 ml/kg, if necessary, to get PP < 30 If VT < 6 ml/kg and PP < 25, increase VT 1 ml/kg (not > 6 ml/kg) If VT = 6 ml/kg, PP < 30, and severe dyspnea, increase VT to 7 or 8 ml/kg (if PP remains < 30) If blood very acidotic, OK for VT and PP to be > 6ml/kg and 30, respectively C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 7/42

23 Controversy: Non-adherence C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 8/42

24 Controversy: Non-adherence ARMA ventilation strategy comparison result: 40% vs. 31% in-hospital mortality (p=0.007) C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 8/42

25 Controversy: Non-adherence ARMA ventilation strategy comparison result: 40% vs. 31% in-hospital mortality (p=0.007) Some argued that the LTVV is not optimal because the adherence to the LTVV protocol was not perfect. C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 8/42

26 Controversy: Non-adherence ARMA ventilation strategy comparison result: 40% vs. 31% in-hospital mortality (p=0.007) Some argued that the LTVV is not optimal because the adherence to the LTVV protocol was not perfect. Physicians deviated from the LTVV protocol when they thought specific aspects of the protocol were detrimental to their patients. For example, when patients appeared to be very dyspneic (short of breath) while receiving 6 ml/kg PBW, some physicians raised tidal volume above 6 ml/kg PBW in an attempt to alleviate the dyspnea. C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 8/42

27 Controversy: Non-adherence ARMA ventilation strategy comparison result: 40% vs. 31% in-hospital mortality (p=0.007) Some argued that the LTVV is not optimal because the adherence to the LTVV protocol was not perfect. Physicians deviated from the LTVV protocol when they thought specific aspects of the protocol were detrimental to their patients. For example, when patients appeared to be very dyspneic (short of breath) while receiving 6 ml/kg PBW, some physicians raised tidal volume above 6 ml/kg PBW in an attempt to alleviate the dyspnea. The deviations from the LTVV protocol may have improved the outcome of some patients. C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 8/42

28 Adherence Specification C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 9/42

29 Adherence Specification In the ARMA trial, VT selection procedure was not recorded C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 9/42

30 Adherence Specification In the ARMA trial, VT selection procedure was not recorded A set of specifications were developed, using VT, PP, PH etc., to determine if LTVV was followed C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 9/42

31 Adherence Specification In the ARMA trial, VT selection procedure was not recorded A set of specifications were developed, using VT, PP, PH etc., to determine if LTVV was followed Plateau pressure > 30 tidal volume 4.5 or ph < 7.2 Tidal volume 3 tidal volume 8.5 ph < 7.2 and set respiratory rate < tidal volume 8.5 (ph < 7.2 and set respiratory rate 35) or (plateau pressure 30 and severe dyspnea) tidal volume 6.5 Plateau pressure 30 or ph < < tidal volume < Plateau pressure tidal volume 4.5 Plateau pressure 25 C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 9/42

32 Adherence Specification In the ARMA trial, VT selection procedure was not recorded A set of specifications were developed, using VT, PP, PH etc., to determine if LTVV was followed Plateau pressure > 30 tidal volume 4.5 or ph < 7.2 Tidal volume 3 tidal volume 8.5 ph < 7.2 and set respiratory rate < tidal volume 8.5 (ph < 7.2 and set respiratory rate 35) or (plateau pressure 30 and severe dyspnea) tidal volume 6.5 Plateau pressure 30 or ph < < tidal volume < Plateau pressure tidal volume 4.5 Plateau pressure 25 Adherence: (1 or 2) AND (3 or 4 or 5 or 6 or 7) C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 9/42

33 Adherence Specification In the ARMA trial, VT selection procedure was not recorded A set of specifications were developed, using VT, PP, PH etc., to determine if LTVV was followed Plateau pressure > 30 tidal volume 4.5 or ph < 7.2 Tidal volume 3 tidal volume 8.5 ph < 7.2 and set respiratory rate < tidal volume 8.5 (ph < 7.2 and set respiratory rate 35) or (plateau pressure 30 and severe dyspnea) tidal volume 6.5 Plateau pressure 30 or ph < < tidal volume < Plateau pressure tidal volume 4.5 Plateau pressure 25 Adherence: (1 or 2) AND (3 or 4 or 5 or 6 or 7) Over 28% observed ventilation days deviated from the LTVV. C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 9/42

34 Goal We seek to draw inference about the distribution of survival that would have resulted had all patients in the LTVV arm of the ARDSNet study strictly adhered to the LTVV protocol. C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 10/42

35 Statistical Challenges C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 11/42

36 Statistical Challenges Time-Dependent Dynamic Treatment Regime C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 11/42

37 Statistical Challenges Time-Dependent Dynamic Treatment Regime In a non-dynamic (static) treatment regime the treatment level and type are specified before treatment is started and not changed with a patient s response C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 11/42

38 Statistical Challenges Time-Dependent Dynamic Treatment Regime In a non-dynamic (static) treatment regime the treatment level and type are specified before treatment is started and not changed with a patient s response LTVV is individually tailored treatment: the treatment is provided to patients only when it is needed and the level of treatment is adjusted based on time-dependent measurements of the patient s need for treatment C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 11/42

39 Statistical Challenges Time-Dependent Dynamic Treatment Regime In a non-dynamic (static) treatment regime the treatment level and type are specified before treatment is started and not changed with a patient s response LTVV is individually tailored treatment: the treatment is provided to patients only when it is needed and the level of treatment is adjusted based on time-dependent measurements of the patient s need for treatment The challenge is to isolate the effect of LTVV from the confounding effect of the time-dependent confounders C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 11/42

40 Statistical Challenges Time-Dependent Dynamic Treatment Regime In a non-dynamic (static) treatment regime the treatment level and type are specified before treatment is started and not changed with a patient s response LTVV is individually tailored treatment: the treatment is provided to patients only when it is needed and the level of treatment is adjusted based on time-dependent measurements of the patient s need for treatment The challenge is to isolate the effect of LTVV from the confounding effect of the time-dependent confounders G-computation C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 11/42

41 Statistical Challenges Time-Dependent Dynamic Treatment Regime In a non-dynamic (static) treatment regime the treatment level and type are specified before treatment is started and not changed with a patient s response LTVV is individually tailored treatment: the treatment is provided to patients only when it is needed and the level of treatment is adjusted based on time-dependent measurements of the patient s need for treatment The challenge is to isolate the effect of LTVV from the confounding effect of the time-dependent confounders G-computation Missing Data C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 11/42

42 Statistical Challenges Time-Dependent Dynamic Treatment Regime In a non-dynamic (static) treatment regime the treatment level and type are specified before treatment is started and not changed with a patient s response LTVV is individually tailored treatment: the treatment is provided to patients only when it is needed and the level of treatment is adjusted based on time-dependent measurements of the patient s need for treatment The challenge is to isolate the effect of LTVV from the confounding effect of the time-dependent confounders G-computation Missing Data Missing of ventilator parameters was non-negligible(e.g. PP 11.6%, arterial ph 28.7%) C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 11/42

43 Statistical Challenges Time-Dependent Dynamic Treatment Regime In a non-dynamic (static) treatment regime the treatment level and type are specified before treatment is started and not changed with a patient s response LTVV is individually tailored treatment: the treatment is provided to patients only when it is needed and the level of treatment is adjusted based on time-dependent measurements of the patient s need for treatment The challenge is to isolate the effect of LTVV from the confounding effect of the time-dependent confounders G-computation Missing Data Missing of ventilator parameters was non-negligible(e.g. PP 11.6%, arterial ph 28.7%) Fully parametric model assuming missing at random C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 11/42

44 Outline 1 Acute Lung Injury and ARMA Trial 2 Bayesian Causal Inference 3 R and BUGS 4 Application Results of the ARMA Trial 5 Concluding Remarks C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 12/42

45 Data Structure For an individual patient, the observed data is as follows: Var Definition and Description Day 1... Day 28 S t Survival HSP t Hospitalization ICU t ICU indicator VNT t Ventilator indicator LC t * Lung compliance PE t Positive end-expiratory pressure PH t ph PD t PP t PEEP t SO t Oxygen saturation SR t Respiratory set rate DS t Severe dyspnea OF t * Organ failure AP Baseline Apache score * OF is observed at baseline. LC = VT/PD C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 13/42

46 Notation C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 14/42

47 Notation Let L t = {S t,hsp t ICU t,vnt t,lc t,pe t,ph t,pd t,so t,sr t,ds t,of t } be the scheduled observed data on day t (t = 0,...,28) C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 14/42

48 Notation Let L t = {S t,hsp t ICU t,vnt t,lc t,pe t,ph t,pd t,so t,sr t,ds t,of t } be the scheduled observed data on day t (t = 0,...,28) Let L t = {L 0,L 1,...,L t } C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 14/42

49 Notation Let L t = {S t,hsp t ICU t,vnt t,lc t,pe t,ph t,pd t,so t,sr t,ds t,of t } be the scheduled observed data on day t (t = 0,...,28) Let L t = {L 0,L 1,...,L t } Let R t = I(VNT t = 1) be the at-risk indicator on day t C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 14/42

50 Notation Let L t = {S t,hsp t ICU t,vnt t,lc t,pe t,ph t,pd t,so t,sr t,ds t,of t } be the scheduled observed data on day t (t = 0,...,28) Let L t = {L 0,L 1,...,L t } Let R t = I(VNT t = 1) be the at-risk indicator on day t Let H t = {HSP t,icu t,vnt t,lc t } and U t = {PE t,ph t,pd t,so t,sr t,ds t,of t } C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 14/42

51 Notation Let L t = {S t,hsp t ICU t,vnt t,lc t,pe t,ph t,pd t,so t,sr t,ds t,of t } be the scheduled observed data on day t (t = 0,...,28) Let L t = {L 0,L 1,...,L t } Let R t = I(VNT t = 1) be the at-risk indicator on day t Let H t = {HSP t,icu t,vnt t,lc t } and U t = {PE t,ph t,pd t,so t,sr t,ds t,of t } Let A t = I((VT t,pp t,ph t,sr t,ds t ) S) where S is the set of combinations defined in the adherence table C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 14/42

52 Notation Let L t = {S t,hsp t ICU t,vnt t,lc t,pe t,ph t,pd t,so t,sr t,ds t,of t } be the scheduled observed data on day t (t = 0,...,28) Let L t = {L 0,L 1,...,L t } Let R t = I(VNT t = 1) be the at-risk indicator on day t Let H t = {HSP t,icu t,vnt t,lc t } and U t = {PE t,ph t,pd t,so t,sr t,ds t,of t } Let A t = I((VT t,pp t,ph t,sr t,ds t ) S) where S is the set of combinations defined in the adherence table Let A t = 1 if S t = 0 or R t = 0 C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 14/42

53 Notation Let L t = {S t,hsp t ICU t,vnt t,lc t,pe t,ph t,pd t,so t,sr t,ds t,of t } be the scheduled observed data on day t (t = 0,...,28) Let L t = {L 0,L 1,...,L t } Let R t = I(VNT t = 1) be the at-risk indicator on day t Let H t = {HSP t,icu t,vnt t,lc t } and U t = {PE t,ph t,pd t,so t,sr t,ds t,of t } Let A t = I((VT t,pp t,ph t,sr t,ds t ) S) where S is the set of combinations defined in the adherence table Let A t = 1 if S t = 0 or R t = 0 Let S t be the counterfactual survival. That is, the survival that would have been observed had a patient been treated according to the LTVV protocol C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 14/42

54 Notation Let L t = {S t,hsp t ICU t,vnt t,lc t,pe t,ph t,pd t,so t,sr t,ds t,of t } be the scheduled observed data on day t (t = 0,...,28) Let L t = {L 0,L 1,...,L t } Let R t = I(VNT t = 1) be the at-risk indicator on day t Let H t = {HSP t,icu t,vnt t,lc t } and U t = {PE t,ph t,pd t,so t,sr t,ds t,of t } Let A t = I((VT t,pp t,ph t,sr t,ds t ) S) where S is the set of combinations defined in the adherence table Let A t = 1 if S t = 0 or R t = 0 Let S t be the counterfactual survival. That is, the survival that would have been observed had a patient been treated according to the LTVV protocol re: The goal is to estimate P(S t = 1) for all t C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 14/42

55 Assumption I C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 15/42

56 Assumption I Stochastic Consistency: P[S t = 1 S t 1 = 1,Āt 1 = 1, H t 1 = h t 1,Ūt 1 = ū t 1 ] = P[S t = 1 S t 1 = 1,Āt 1 = 1, H t 1 = h t 1,Ūt 1 = ū t 1 ] This assumption implies that the potential survival outcome under the LTVV protocol is uniquely determined. That is, if the physician followed the protocol, the treatment for an individual is unique and therefore her potential survival outcome is unique. C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 15/42

57 Assumption II Evaluable Outcome P[A t = 1 S t = 1,Āt 1 = 1, H t = h t,ūt 1 = ū t 1 ] > 0 This assumes that when a group of patients have been treated with the LTVV protocol until day t, at least some of them will continue to be treated with adherence to the protocol on day t+1. C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 16/42

58 Assumption III Sequential Randomization A t (S k : k = 2,...,28) Ā t 1,S t = 1,R t = 1, H t,ūt 1 This assumption will hold if H t and Ūt 1 includes all risk factors, other than treatment history, used by the physician to determine the patient s ventilator settings on day t. C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 17/42

59 G-computation C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 18/42

60 G-computation Given the distribution of L 28 and the three assumptions above, the potential survivor function at day m (m 28) is identified via the following G-computation formula: C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 18/42

61 G-computation Given the distribution of L 28 and the three assumptions above, the potential survivor function at day m (m 28) is identified via the following G-computation formula: P[S m = 1] = P[Sm = 1 S 1 = 1, H 1 = h 1,U 0 = u 0 ]df(h 1 S 1 = 1,U 0 = u 0,H 0 = h 0 ) h 0 u 0 h 1 P[S 1 = 1 U 0 = u 0,H 0 = h 0 ]df(u 0 H 0 = h 0 )df(h 0 ) = P[Sm = 1 S 1 = 1,A 1 = 1, H 1 = h 1,U 0 = u 0 ] h 0 u 0 h 1 }{{} sequential randomization df(h 1 S 1 = 1,U 0 = u 0,H 0 = h 0 ) P[S 1 = 1 U 0 = u 0,H 0 = h 0 ]df(u 0 H 0 = h 0 )df(h 0 ) C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 18/42

62 G-computation = P[Sm = 1 S 2 = 1,A 1 = 1, H 2 = h 2,Ū1 = ū 1 ] h 0 u 0 h 1 u 1 h 2 df(h 2 S 2 = 1,A 1 = 1, H 1 = h 1,Ū1 = ū 1 ) P[S 2 = 1 S 1 = 1,A 1 = 1, H 1 = h 1,Ū1 = ū 1 ] df(u 1 S 1 = 1,A 1 = 1, H 1 = h 1,U 0 = u 0 ) df(h 1 S 1 = 1,U 0 = u 0,H 0 = h 0 ) P[S 1 = 1 U 0 = u 0,H 0 = h 0 ]df(u 0 H 0 = h 0 )df(h 0 ) = P[Sm = 1 S 2 = 1,Ā2 = 1, H 2 = h 2,Ū1 = ū 1 ]... h 0 u 0 h 1 u 1 h 2 }{{} sequential randomization C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 19/42

63 G-computation. =... P[Sm = 1 S m 1 = 1,Ām 1 = 1, H m 1 = h m 1,Ūm 1 = ū m 1 ] h 0 u 0 h 1 u 1 h m 1 u m 1 m 1 t=1 { P[S t = 1 S t 1 = 1,Āt 1 = 1, H t 1 = h t 1,Ūt 1 = ū t 1 ] df(h t S t = 1,Āt 1 = 1, H t 1 = h t 1,Ūt 1 = ū t 1 ) } df(u t S t = 1,Āt = 1, H t = h t,ūt 1 = ū t 1 ) df(u 0 H 0 = h 0 )df(h 0 ) =... P[S m = 1 S m 1 = 1,Ām 1 = 1, H m 1 = h m 1,Ūm 1 = ū m 1 ] h 0 u 0 h 1 u 1 h m 1 u m 1 }{{} stochastic consistency... C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 20/42

64 Modeling of L 28 The likelihood of L 28 is as follows: Lik = df(h 0 )df(u 0 H 0 ) 28 t=1 { df(u t H t,ūt 1,S t = 1) df(h t H t 1,Ūt 1,S t = 1)} St df(st S t 1 = 1, H t 1,Ūt 1) S t 1. Additional Markovian Assumption Lik = df(h 0 )df(u 0 H 0 ) 28 t=1 { df(u t H t,u t 1,H 0,S t = 1) df(h t H t 1,U t 1,H 0,S t = 1)} StdF(St S t 1 = 1,H t 1,U t 1,H 0 ) S t 1. That is,a patient s status on day t only depends on his most recent history and the baseline H 0. C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 21/42

65 Models of S t df(s t S t 1 = 1,H t 1,U t 1,H 0 ) = df(s t S t 1 = 1,R t 1 = 1,G t 1,U t 1,H 0 ) Rt 1 df(s t S t 1 = 1,HSP t 1 = 1,ICU t 1 = 0,OF t 1,H 0 ) (1 Rt 1)(1 ICUt 1) ICUt 1(1 V NTt 1) df(s t S t 1 = 1,HSP t 1 = 1,ICU t 1 = 1,VNT t 1 = 0,OF t 1,H 0 ) for t = 1,...,28 C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 22/42

66 Models of H t df(h t S t = 1,H t 1,U t 1,H 0 ) = df(hsp t S t = 1,HSP t 1 = 1,ICU t 1 = 0,OF t 1,H 0 ) HSPt 1(1 ICUt 1) df(icu t S t = 1,HSP t = 1,VNT t 1 = 0,OF t 1,H 0 ) HSPt(1 VNTt 1) df(vnt t S t = 1,HSP t = 1,ICU t = 1,R t 1 = 1,G t 1,U t 1,H 0 ) HSPtICUtRt 1 df(vnt t S t = 1,HSP t = 1,ICU t = 1,R t 1 = 0,OF t 1,H 0 ) HSPtICUt(1 Rt 1) df(lc t S t = 1,R t = 1,R t 1 = 1,G t 1,U t 1,H 0 ) RtRt 1 df(lc t S t = 1,R t = 1,R t 1 = 0,OF t 1,H 0 ) Rt(1 Rt 1) df(pe t S t = 1,R t = 1,R t 1 = 1,LC t,pe t 1,PD t 1,DS t 1,OF t 1,H 0 ) RtRt 1 df(pe t S t = 1,R t = 1,R t 1 = 0,LC t,of t 1,H 0 ) Rt(1 Rt 1) for t = 1,...,28 C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 23/42

67 Models of U t df(ut St = 1,Ht,Ut 1,H0) = df(pht St = 1,Rt = 1,Rt 1 = 1,LCt,PHt 1,DSt 1,OFt 1,H0) RtRt 1 df(pht St = 1,Rt = 1,Rt 1 = 0,LCt,OFt 1,H0) Rt(1 Rt 1) df(pdt St = 1,Rt = 1,Rt 1 = 1,LCt,PEt,PDt 1,H0) RtRt 1 df(pdt St = 1,Rt = 1,Rt 1 = 0,LCt,PEt,OFt 1,H0) Rt(1 Rt 1) df(sot St = 1,Rt = 1,Rt 1 = 1,LCt,PEt,PHt,PDt,SOt 1,H0) RtRt 1 df(sot St = 1,Rt = 1,Rt 1 = 0,LCt,PEt,PHt,PDt,OFt 1,H0) Rt(1 Rt 1) df(srt St = 1,Rt = 1,Rt 1 = 1,LCt,PEt,PHt,SRt 1,DSt 1,OFt 1,H0) RtRt 1 df(srt St = 1,Rt = 1,Rt 1 = 0,LCt,PEt,PHt,OFt 1,H0) Rt(1 Rt 1) df(dst St = 1,Rt = 1,Rt 1 = 1,LCt,PEt,PHt,SOt,SRt,DSt 1,OFt 1,H0) RtRt 1 df(dst St = 1,Rt = 1,Rt 1 = 0,LCt,PEt,PHt,,SOt,SRt,OFt 1,H0) Rt(1 Rt 1) df(oft St = 1,Rt = 1,LCt,PEt,PHt,SOt,SRt,DSt,OFt 1,H0) Rt df(oft St = 1,Rt = 0,Rt 1 = 1,LCt 1,PEt 1,PHt 1,SOt 1,SRt 1,DSt 1,OFt 1,H0) (1 Rt)Rt 1 df(oft St = 1,HSPt = 1,VNTt = 0,Rt 1 = 0,ICUt 1 = 1,OFt 1,H0) HSPt(1 VNTt)(1 V NTt 1)ICUt 1 df(oft St = 1,HSPt = 1,VNTt = 0,HSPt 1 = 1,ICUt 1 = 0,OFt 1,H0) HSPt(1 VNTt)HSPt 1(1 ICUt 1) for t = 1,...,28 C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 24/42

68 Monte Carlo Integration Algorithm C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 25/42

69 Monte Carlo Integration Algorithm (a) Set t = 0; sample AP, OF 0 ; set p 0 = 1 C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 25/42

70 Monte Carlo Integration Algorithm (a) Set t = 0; sample AP, OF 0 ; set p 0 = 1 (b) Set t = t+1; set p t = p t 1 P[S t = 1 S t 1 = 1,H t 1,U t 1,H 0 ]. If t = 28, stop C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 25/42

71 Monte Carlo Integration Algorithm (a) Set t = 0; sample AP, OF 0 ; set p 0 = 1 (b) Set t = t+1; set p t = p t 1 P[S t = 1 S t 1 = 1,H t 1,U t 1,H 0 ]. If t = 28, stop (c) Sample HSP t, ICU t, VNT t, LC t C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 25/42

72 Monte Carlo Integration Algorithm (a) Set t = 0; sample AP, OF 0 ; set p 0 = 1 (b) Set t = t+1; set p t = p t 1 P[S t = 1 S t 1 = 1,H t 1,U t 1,H 0 ]. If t = 28, stop (c) Sample HSP t, ICU t, VNT t, LC t (d) Sample PE t, PH t, PD t, SO t, DS t C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 25/42

73 Monte Carlo Integration Algorithm (a) Set t = 0; sample AP, OF 0 ; set p 0 = 1 (b) Set t = t+1; set p t = p t 1 P[S t = 1 S t 1 = 1,H t 1,U t 1,H 0 ]. If t = 28, stop (c) Sample HSP t, ICU t, VNT t, LC t (d) Sample PE t, PH t, PD t, SO t, DS t (e) Set A t = I((VT t,sr t,ph t,pp t,ds t ) S). If A t = 0, go to (d) C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 25/42

74 Monte Carlo Integration Algorithm (a) Set t = 0; sample AP, OF 0 ; set p 0 = 1 (b) Set t = t+1; set p t = p t 1 P[S t = 1 S t 1 = 1,H t 1,U t 1,H 0 ]. If t = 28, stop (c) Sample HSP t, ICU t, VNT t, LC t (d) Sample PE t, PH t, PD t, SO t, DS t (e) Set A t = I((VT t,sr t,ph t,pp t,ds t ) S). If A t = 0, go to (d) (f) Sample OF t ; go to (b) C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 25/42

75 Outline 1 Acute Lung Injury and ARMA Trial 2 Bayesian Causal Inference 3 R and BUGS 4 Application Results of the ARMA Trial 5 Concluding Remarks C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 26/42

76 Flowchart Observed Data Data Manipulation in R R Calls BUGS BUGS Script Processed Data Initial Values Priors BUGS Model BUGS MCMC Sampling Posterior Samples N R Parallel G-computation Jobs N 4 R Summarization Results C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 27/42

77 The BUGS Project C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 28/42

78 The BUGS Project There are several versions of BUGS, which can be confusing C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 28/42

79 The BUGS Project There are several versions of BUGS, which can be confusing WinBUGS C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 28/42

80 The BUGS Project There are several versions of BUGS, which can be confusing WinBUGS Latest version C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 28/42

81 The BUGS Project There are several versions of BUGS, which can be confusing WinBUGS Latest version Stand-alone and considered stable C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 28/42

82 The BUGS Project There are several versions of BUGS, which can be confusing WinBUGS Latest version Stand-alone and considered stable Can be extended. Add-ons: GeoBUGS, PKBUGS C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 28/42

83 The BUGS Project There are several versions of BUGS, which can be confusing WinBUGS Latest version Stand-alone and considered stable Can be extended. Add-ons: GeoBUGS, PKBUGS Can be installed in Linux via wine and virtual screen server (e.g. Xvfb) C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 28/42

84 The BUGS Project There are several versions of BUGS, which can be confusing WinBUGS Latest version Stand-alone and considered stable Can be extended. Add-ons: GeoBUGS, PKBUGS Can be installed in Linux via wine and virtual screen server (e.g. Xvfb) OpenBUGS C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 28/42

85 The BUGS Project There are several versions of BUGS, which can be confusing WinBUGS Latest version Stand-alone and considered stable Can be extended. Add-ons: GeoBUGS, PKBUGS Can be installed in Linux via wine and virtual screen server (e.g. Xvfb) OpenBUGS Latest version C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 28/42

86 The BUGS Project There are several versions of BUGS, which can be confusing WinBUGS Latest version Stand-alone and considered stable Can be extended. Add-ons: GeoBUGS, PKBUGS Can be installed in Linux via wine and virtual screen server (e.g. Xvfb) OpenBUGS Latest version LinBUGS is the Linux based OpenBUGS C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 28/42

87 The BUGS Project There are several versions of BUGS, which can be confusing WinBUGS Latest version Stand-alone and considered stable Can be extended. Add-ons: GeoBUGS, PKBUGS Can be installed in Linux via wine and virtual screen server (e.g. Xvfb) OpenBUGS Latest version LinBUGS is the Linux based OpenBUGS JAGS C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 28/42

88 The BUGS Project There are several versions of BUGS, which can be confusing WinBUGS Latest version Stand-alone and considered stable Can be extended. Add-ons: GeoBUGS, PKBUGS Can be installed in Linux via wine and virtual screen server (e.g. Xvfb) OpenBUGS Latest version LinBUGS is the Linux based OpenBUGS JAGS Just Another Gibbs Sampler, not part of the BUGS project C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 28/42

89 BUGS Script Example script.txt: modelcheck("model.txt") modeldata("data.txt") modelcompile() modelinits("inits.txt") modelupdate(10000) summaryset("a") samplesset("a") modelupdate(10000) summarystats("*") samplesstats("*") samplescoda("*","output") modelquit() C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 29/42

90 BUGS Model Example model.txt: model { for( i in 1 : N ) { mu[i] <- a+b*age[i] price[i] ~ dnorm(mu[i], tau) } tau ~ dgamma(0.001, 0.001) sigma <- 1 / sqrt(tau) a ~ dnorm(0, 1.0E-12) b ~ dnorm(0, 1.0E-12) } C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 30/42

91 BUGS Data Example init.txt: list(a=0, b=0, tau=1) Example data.txt: list(n=9, age = c(13, 14, 14,12, 9, 15, 10, 14, 9), price = c(295, 230, 390, 280, 500, 299, 390, 299, 450)) C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 31/42

92 R Talks to BUGS C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 32/42

93 R Talks to BUGS R2WinBUGS, rbugs C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 32/42

94 R Talks to BUGS R2WinBUGS, rbugs Prepare files needed for running BUGS in batch-mode C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 32/42

95 R Talks to BUGS R2WinBUGS, rbugs Prepare files needed for running BUGS in batch-mode Running BUGS from R C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 32/42

96 R Talks to BUGS R2WinBUGS, rbugs Prepare files needed for running BUGS in batch-mode Running BUGS from R Save posterior samples for easy access in R C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 32/42

97 R Talks to BUGS R2WinBUGS, rbugs Prepare files needed for running BUGS in batch-mode Running BUGS from R Save posterior samples for easy access in R BRugs C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 32/42

98 R Talks to BUGS R2WinBUGS, rbugs Prepare files needed for running BUGS in batch-mode Running BUGS from R Save posterior samples for easy access in R BRugs Part of the OpenBUGS project C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 32/42

99 R Talks to BUGS R2WinBUGS, rbugs Prepare files needed for running BUGS in batch-mode Running BUGS from R Save posterior samples for easy access in R BRugs Part of the OpenBUGS project Instead of generating script file, capable of working interactively with BUGS C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 32/42

100 R Talks to BUGS R2WinBUGS, rbugs Prepare files needed for running BUGS in batch-mode Running BUGS from R Save posterior samples for easy access in R BRugs Part of the OpenBUGS project Instead of generating script file, capable of working interactively with BUGS Not for Linux. A new version may be out in one month or two to support Linux C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 32/42

101 Parallel Computing C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 33/42

102 Parallel Computing High-performance computing (HPC) uses supercomputers and computer clusters to solve advanced computation problems C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 33/42

103 Parallel Computing High-performance computing (HPC) uses supercomputers and computer clusters to solve advanced computation problems Parallel computing can be employed in HPC C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 33/42

104 Parallel Computing High-performance computing (HPC) uses supercomputers and computer clusters to solve advanced computation problems Parallel computing can be employed in HPC Embarrassingly parallel is one parallel computing case where no dependency exists between the parallel tasks C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 33/42

105 Parallel Computing High-performance computing (HPC) uses supercomputers and computer clusters to solve advanced computation problems Parallel computing can be employed in HPC Embarrassingly parallel is one parallel computing case where no dependency exists between the parallel tasks R Parallel Computing Frameworks C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 33/42

106 Parallel Computing High-performance computing (HPC) uses supercomputers and computer clusters to solve advanced computation problems Parallel computing can be employed in HPC Embarrassingly parallel is one parallel computing case where no dependency exists between the parallel tasks R Parallel Computing Frameworks Simple Network of Workstations (SNOW) C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 33/42

107 Parallel Computing High-performance computing (HPC) uses supercomputers and computer clusters to solve advanced computation problems Parallel computing can be employed in HPC Embarrassingly parallel is one parallel computing case where no dependency exists between the parallel tasks R Parallel Computing Frameworks Simple Network of Workstations (SNOW) Simple Parallel R INTerface (SPRINT) C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 33/42

108 Parallel Computing High-performance computing (HPC) uses supercomputers and computer clusters to solve advanced computation problems Parallel computing can be employed in HPC Embarrassingly parallel is one parallel computing case where no dependency exists between the parallel tasks R Parallel Computing Frameworks Simple Network of Workstations (SNOW) Simple Parallel R INTerface (SPRINT) Rmpi C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 33/42

109 Example using Sun Grid Engine Shell Script qsub -N bugs qsub -N gcomp qsub -N rsum bugs.sh -hold_jid bugs -t gcomp.sh -hold_jid gcomp rsum.sh gcomp.sh R --vanilla --args $SGE_TASK_ID < GCOMP.R C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 34/42

110 Outline 1 Acute Lung Injury and ARMA Trial 2 Bayesian Causal Inference 3 R and BUGS 4 Application Results of the ARMA Trial 5 Concluding Remarks C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 35/42

111 Survival Rates Empirical (gray), posterior mean observed (dashed)) and potential (black) survival curves. Shaded area denote 95% posterior credible band for the potential survival curve: Survival Rate Observed Survival (Empirical) Observed Survival (Model Based) Potential Survival (Model Based) Days C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 36/42

112 Survival Rates Difference Posterior mean difference between observed and potential survival curves, with 95% posterior credible band: Difference Days C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 37/42

113 Outline 1 Acute Lung Injury and ARMA Trial 2 Bayesian Causal Inference 3 R and BUGS 4 Application Results of the ARMA Trial 5 Concluding Remarks C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 38/42

114 Summary C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 39/42

115 Summary We used the G-computation formula to estimate the potential survival distribution under the dynamic LTVV treatment regime for patients with acute lung injury C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 39/42

116 Summary We used the G-computation formula to estimate the potential survival distribution under the dynamic LTVV treatment regime for patients with acute lung injury Our results demonstrate that strict adherence to the LTVV algorithm does not lead to increased mortality C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 39/42

117 Summary We used the G-computation formula to estimate the potential survival distribution under the dynamic LTVV treatment regime for patients with acute lung injury Our results demonstrate that strict adherence to the LTVV algorithm does not lead to increased mortality The current methodology relies on untestable assumption that all the time-dependent confounders have been measured, i.e. no unmeasured confounding. Sensitivity of the conclusion to deviations from this assumption should be explored C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 39/42

118 Summary We used the G-computation formula to estimate the potential survival distribution under the dynamic LTVV treatment regime for patients with acute lung injury Our results demonstrate that strict adherence to the LTVV algorithm does not lead to increased mortality The current methodology relies on untestable assumption that all the time-dependent confounders have been measured, i.e. no unmeasured confounding. Sensitivity of the conclusion to deviations from this assumption should be explored We illustrated the utility of R and OpenBUGS for the Bayesian implementation of the proposed methodology C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 39/42

119 Summary We used the G-computation formula to estimate the potential survival distribution under the dynamic LTVV treatment regime for patients with acute lung injury Our results demonstrate that strict adherence to the LTVV algorithm does not lead to increased mortality The current methodology relies on untestable assumption that all the time-dependent confounders have been measured, i.e. no unmeasured confounding. Sensitivity of the conclusion to deviations from this assumption should be explored We illustrated the utility of R and OpenBUGS for the Bayesian implementation of the proposed methodology With the dissemination and popularization of High Performance Computing, parallel computation will become mainstream for all scientific researchers including statisticians C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 39/42

120 THE END C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 40/42

121 BLANK C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 41/42

122 Tips for R Installation C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 42/42

123 Tips for R Installation R version has been released on 05/31/2010 C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 42/42

124 Tips for R Installation R version has been released on 05/31/2010 Use./configure -prefix= to specify the destination folder for installation C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 42/42

125 Tips for R Installation R version has been released on 05/31/2010 Use./configure -prefix= to specify the destination folder for installation Use export R_LIBS= to set up the local R package library path C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 42/42

126 Tips for R Installation R version has been released on 05/31/2010 Use./configure -prefix= to specify the destination folder for installation Use export R_LIBS= to set up the local R package library path To install R2WinBUGS in the latest version R, packages lattice and coda need to be installed separately C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 42/42

127 Tips for R Installation R version has been released on 05/31/2010 Use./configure -prefix= to specify the destination folder for installation Use export R_LIBS= to set up the local R package library path To install R2WinBUGS in the latest version R, packages lattice and coda need to be installed separately Set export BUGS= for rbugs C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 42/42

128 Tips for R Installation R version has been released on 05/31/2010 Use./configure -prefix= to specify the destination folder for installation Use export R_LIBS= to set up the local R package library path To install R2WinBUGS in the latest version R, packages lattice and coda need to be installed separately Set export BUGS= for rbugs Extensions C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 42/42

129 Tips for R Installation R version has been released on 05/31/2010 Use./configure -prefix= to specify the destination folder for installation Use export R_LIBS= to set up the local R package library path To install R2WinBUGS in the latest version R, packages lattice and coda need to be installed separately Set export BUGS= for rbugs Extensions Use Rprof and summaryrprof to find the computation bottleneck C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 42/42

130 Tips for R Installation R version has been released on 05/31/2010 Use./configure -prefix= to specify the destination folder for installation Use export R_LIBS= to set up the local R package library path To install R2WinBUGS in the latest version R, packages lattice and coda need to be installed separately Set export BUGS= for rbugs Extensions Use Rprof and summaryrprof to find the computation bottleneck Use C or Fortran extensions to expedite the computation C. Wang, D.O. Scharfstein, M.J. Daniels Evaluate the Causal Effect of DTR 42/42

Advanced Statistical Modelling

Advanced Statistical Modelling Markov chain Monte Carlo (MCMC) Methods and Their Applications in Bayesian Statistics School of Technology and Business Studies/Statistics Dalarna University Borlänge, Sweden. Feb. 05, 2014. Outlines 1

More information

Robust Bayesian Regression

Robust Bayesian Regression Readings: Hoff Chapter 9, West JRSSB 1984, Fúquene, Pérez & Pericchi 2015 Duke University November 17, 2016 Body Fat Data: Intervals w/ All Data Response % Body Fat and Predictor Waist Circumference 95%

More information

Introduction to Bayesian Statistics and Markov Chain Monte Carlo Estimation. EPSY 905: Multivariate Analysis Spring 2016 Lecture #10: April 6, 2016

Introduction to Bayesian Statistics and Markov Chain Monte Carlo Estimation. EPSY 905: Multivariate Analysis Spring 2016 Lecture #10: April 6, 2016 Introduction to Bayesian Statistics and Markov Chain Monte Carlo Estimation EPSY 905: Multivariate Analysis Spring 2016 Lecture #10: April 6, 2016 EPSY 905: Intro to Bayesian and MCMC Today s Class An

More information

WinBUGS : part 2. Bruno Boulanger Jonathan Jaeger Astrid Jullion Philippe Lambert. Gabriele, living with rheumatoid arthritis

WinBUGS : part 2. Bruno Boulanger Jonathan Jaeger Astrid Jullion Philippe Lambert. Gabriele, living with rheumatoid arthritis WinBUGS : part 2 Bruno Boulanger Jonathan Jaeger Astrid Jullion Philippe Lambert Gabriele, living with rheumatoid arthritis Agenda 2! Hierarchical model: linear regression example! R2WinBUGS Linear Regression

More information

Prerequisite: STATS 7 or STATS 8 or AP90 or (STATS 120A and STATS 120B and STATS 120C). AP90 with a minimum score of 3

Prerequisite: STATS 7 or STATS 8 or AP90 or (STATS 120A and STATS 120B and STATS 120C). AP90 with a minimum score of 3 University of California, Irvine 2017-2018 1 Statistics (STATS) Courses STATS 5. Seminar in Data Science. 1 Unit. An introduction to the field of Data Science; intended for entering freshman and transfers.

More information

MASSACHUSETTS INSTITUTE OF TECHNOLOGY

MASSACHUSETTS INSTITUTE OF TECHNOLOGY Harvard-MIT Division of Health Sciences and Technology HST.54J: Quantitative Physiology: Organ Transport Systems Instructors: Roger Mark and Jose Venegas MASSACHUSETTS INSTITUTE OF TECHNOLOGY Departments

More information

Using Post Outcome Measurement Information in Censoring by Death Problems

Using Post Outcome Measurement Information in Censoring by Death Problems Using Post Outcome Measurement Information in Censoring by Death Problems Fan Yang University of Chicago, Chicago, USA. Dylan S. Small University of Pennsylvania, Philadelphia, USA. Summary. Many clinical

More information

Targeted Maximum Likelihood Estimation in Safety Analysis

Targeted Maximum Likelihood Estimation in Safety Analysis Targeted Maximum Likelihood Estimation in Safety Analysis Sam Lendle 1 Bruce Fireman 2 Mark van der Laan 1 1 UC Berkeley 2 Kaiser Permanente ISPE Advanced Topics Session, Barcelona, August 2012 1 / 35

More information

Bayesian Statistics Comparing Two Proportions or Means

Bayesian Statistics Comparing Two Proportions or Means Bayesian Statistics Comparing Two Proportions or Means Michael Anderson, PhD Hélène Carabin, DVM, PhD Department of Biostatistics and Epidemiology The University of Oklahoma Health Sciences Center May

More information

A multi-state model for the prognosis of non-mild acute pancreatitis

A multi-state model for the prognosis of non-mild acute pancreatitis A multi-state model for the prognosis of non-mild acute pancreatitis Lore Zumeta Olaskoaga 1, Felix Zubia Olaskoaga 2, Guadalupe Gómez Melis 1 1 Universitat Politècnica de Catalunya 2 Intensive Care Unit,

More information

The STS Surgeon Composite Technical Appendix

The STS Surgeon Composite Technical Appendix The STS Surgeon Composite Technical Appendix Overview Surgeon-specific risk-adjusted operative operative mortality and major complication rates were estimated using a bivariate random-effects logistic

More information

Causal Hazard Ratio Estimation By Instrumental Variables or Principal Stratification. Todd MacKenzie, PhD

Causal Hazard Ratio Estimation By Instrumental Variables or Principal Stratification. Todd MacKenzie, PhD Causal Hazard Ratio Estimation By Instrumental Variables or Principal Stratification Todd MacKenzie, PhD Collaborators A. James O Malley Tor Tosteson Therese Stukel 2 Overview 1. Instrumental variable

More information

Case Study in the Use of Bayesian Hierarchical Modeling and Simulation for Design and Analysis of a Clinical Trial

Case Study in the Use of Bayesian Hierarchical Modeling and Simulation for Design and Analysis of a Clinical Trial Case Study in the Use of Bayesian Hierarchical Modeling and Simulation for Design and Analysis of a Clinical Trial William R. Gillespie Pharsight Corporation Cary, North Carolina, USA PAGE 2003 Verona,

More information

A Bayesian Nonparametric Approach to Monotone Missing Data in Longitudinal Studies with Informative Missingness

A Bayesian Nonparametric Approach to Monotone Missing Data in Longitudinal Studies with Informative Missingness A Bayesian Nonparametric Approach to Monotone Missing Data in Longitudinal Studies with Informative Missingness A. Linero and M. Daniels UF, UT-Austin SRC 2014, Galveston, TX 1 Background 2 Working model

More information

Bayesian Graphical Models

Bayesian Graphical Models Graphical Models and Inference, Lecture 16, Michaelmas Term 2009 December 4, 2009 Parameter θ, data X = x, likelihood L(θ x) p(x θ). Express knowledge about θ through prior distribution π on θ. Inference

More information

A Bayesian Nonparametric Approach to Causal Inference for Semi-competing risks

A Bayesian Nonparametric Approach to Causal Inference for Semi-competing risks A Bayesian Nonparametric Approach to Causal Inference for Semi-competing risks Y. Xu, D. Scharfstein, P. Mueller, M. Daniels Johns Hopkins, Johns Hopkins, UT-Austin, UF JSM 2018, Vancouver 1 What are semi-competing

More information

Causal Inference Basics

Causal Inference Basics Causal Inference Basics Sam Lendle October 09, 2013 Observed data, question, counterfactuals Observed data: n i.i.d copies of baseline covariates W, treatment A {0, 1}, and outcome Y. O i = (W i, A i,

More information

Interim Monitoring of Clinical Trials: Decision Theory, Dynamic Programming. and Optimal Stopping

Interim Monitoring of Clinical Trials: Decision Theory, Dynamic Programming. and Optimal Stopping Interim Monitoring of Clinical Trials: Decision Theory, Dynamic Programming and Optimal Stopping Christopher Jennison Department of Mathematical Sciences, University of Bath, UK http://people.bath.ac.uk/mascj

More information

Estimating Causal Effects of Organ Transplantation Treatment Regimes

Estimating Causal Effects of Organ Transplantation Treatment Regimes Estimating Causal Effects of Organ Transplantation Treatment Regimes David M. Vock, Jeffrey A. Verdoliva Boatman Division of Biostatistics University of Minnesota July 31, 2018 1 / 27 Hot off the Press

More information

DIC: Deviance Information Criterion

DIC: Deviance Information Criterion (((( Welcome Page Latest News DIC: Deviance Information Criterion Contact us/bugs list WinBUGS New WinBUGS examples FAQs DIC GeoBUGS DIC (Deviance Information Criterion) is a Bayesian method for model

More information

Estimating the Mean Response of Treatment Duration Regimes in an Observational Study. Anastasios A. Tsiatis.

Estimating the Mean Response of Treatment Duration Regimes in an Observational Study. Anastasios A. Tsiatis. Estimating the Mean Response of Treatment Duration Regimes in an Observational Study Anastasios A. Tsiatis http://www.stat.ncsu.edu/ tsiatis/ Introduction to Dynamic Treatment Regimes 1 Outline Description

More information

A multi-state model for the prognosis of non-mild acute pancreatitis

A multi-state model for the prognosis of non-mild acute pancreatitis A multi-state model for the prognosis of non-mild acute pancreatitis Lore Zumeta Olaskoaga 1, Felix Zubia Olaskoaga 2, Marta Bofill Roig 1, Guadalupe Gómez Melis 1 1 Universitat Politècnica de Catalunya

More information

Bayesian philosophy Bayesian computation Bayesian software. Bayesian Statistics. Petter Mostad. Chalmers. April 6, 2017

Bayesian philosophy Bayesian computation Bayesian software. Bayesian Statistics. Petter Mostad. Chalmers. April 6, 2017 Chalmers April 6, 2017 Bayesian philosophy Bayesian philosophy Bayesian statistics versus classical statistics: War or co-existence? Classical statistics: Models have variables and parameters; these are

More information

Causal Inference. Miguel A. Hernán, James M. Robins. May 19, 2017

Causal Inference. Miguel A. Hernán, James M. Robins. May 19, 2017 Causal Inference Miguel A. Hernán, James M. Robins May 19, 2017 ii Causal Inference Part III Causal inference from complex longitudinal data Chapter 19 TIME-VARYING TREATMENTS So far this book has dealt

More information

Optimising Group Sequential Designs. Decision Theory, Dynamic Programming. and Optimal Stopping

Optimising Group Sequential Designs. Decision Theory, Dynamic Programming. and Optimal Stopping : Decision Theory, Dynamic Programming and Optimal Stopping Christopher Jennison Department of Mathematical Sciences, University of Bath, UK http://people.bath.ac.uk/mascj InSPiRe Conference on Methodology

More information

BUGS Bayesian inference Using Gibbs Sampling

BUGS Bayesian inference Using Gibbs Sampling BUGS Bayesian inference Using Gibbs Sampling Glen DePalma Department of Statistics May 30, 2013 www.stat.purdue.edu/~gdepalma 1 / 20 Bayesian Philosophy I [Pearl] turned Bayesian in 1971, as soon as I

More information

Individualized Treatment Effects with Censored Data via Nonparametric Accelerated Failure Time Models

Individualized Treatment Effects with Censored Data via Nonparametric Accelerated Failure Time Models Individualized Treatment Effects with Censored Data via Nonparametric Accelerated Failure Time Models Nicholas C. Henderson Thomas A. Louis Gary Rosner Ravi Varadhan Johns Hopkins University July 31, 2018

More information

Adaptive Prediction of Event Times in Clinical Trials

Adaptive Prediction of Event Times in Clinical Trials Adaptive Prediction of Event Times in Clinical Trials Yu Lan Southern Methodist University Advisor: Daniel F. Heitjan May 8, 2017 Yu Lan (SMU) May 8, 2017 1 / 19 Clinical Trial Prediction Event-based trials:

More information

Probabilistic Machine Learning

Probabilistic Machine Learning Probabilistic Machine Learning Bayesian Nets, MCMC, and more Marek Petrik 4/18/2017 Based on: P. Murphy, K. (2012). Machine Learning: A Probabilistic Perspective. Chapter 10. Conditional Independence Independent

More information

Marginal versus conditional effects: does it make a difference? Mireille Schnitzer, PhD Université de Montréal

Marginal versus conditional effects: does it make a difference? Mireille Schnitzer, PhD Université de Montréal Marginal versus conditional effects: does it make a difference? Mireille Schnitzer, PhD Université de Montréal Overview In observational and experimental studies, the goal may be to estimate the effect

More information

DAG models and Markov Chain Monte Carlo methods a short overview

DAG models and Markov Chain Monte Carlo methods a short overview DAG models and Markov Chain Monte Carlo methods a short overview Søren Højsgaard Institute of Genetics and Biotechnology University of Aarhus August 18, 2008 Printed: August 18, 2008 File: DAGMC-Lecture.tex

More information

A Sampling of IMPACT Research:

A Sampling of IMPACT Research: A Sampling of IMPACT Research: Methods for Analysis with Dropout and Identifying Optimal Treatment Regimes Marie Davidian Department of Statistics North Carolina State University http://www.stat.ncsu.edu/

More information

Causal Sensitivity Analysis for Decision Trees

Causal Sensitivity Analysis for Decision Trees Causal Sensitivity Analysis for Decision Trees by Chengbo Li A thesis presented to the University of Waterloo in fulfillment of the thesis requirement for the degree of Master of Mathematics in Computer

More information

Bayesian Inference and MCMC

Bayesian Inference and MCMC Bayesian Inference and MCMC Aryan Arbabi Partly based on MCMC slides from CSC412 Fall 2018 1 / 18 Bayesian Inference - Motivation Consider we have a data set D = {x 1,..., x n }. E.g each x i can be the

More information

Fast Likelihood-Free Inference via Bayesian Optimization

Fast Likelihood-Free Inference via Bayesian Optimization Fast Likelihood-Free Inference via Bayesian Optimization Michael Gutmann https://sites.google.com/site/michaelgutmann University of Helsinki Aalto University Helsinki Institute for Information Technology

More information

R Demonstration ANCOVA

R Demonstration ANCOVA R Demonstration ANCOVA Objective: The purpose of this week s session is to demonstrate how to perform an analysis of covariance (ANCOVA) in R, and how to plot the regression lines for each level of the

More information

Introduction to Statistical Inference

Introduction to Statistical Inference Introduction to Statistical Inference Kosuke Imai Princeton University January 31, 2010 Kosuke Imai (Princeton) Introduction to Statistical Inference January 31, 2010 1 / 21 What is Statistics? Statistics

More information

Estimating Optimal Dynamic Treatment Regimes from Clustered Data

Estimating Optimal Dynamic Treatment Regimes from Clustered Data Estimating Optimal Dynamic Treatment Regimes from Clustered Data Bibhas Chakraborty Department of Biostatistics, Columbia University bc2425@columbia.edu Society for Clinical Trials Annual Meetings Boston,

More information

2 Describing Contingency Tables

2 Describing Contingency Tables 2 Describing Contingency Tables I. Probability structure of a 2-way contingency table I.1 Contingency Tables X, Y : cat. var. Y usually random (except in a case-control study), response; X can be random

More information

Why Bayesian approaches? The average height of a rare plant

Why Bayesian approaches? The average height of a rare plant Why Bayesian approaches? The average height of a rare plant Estimation and comparison of averages is an important step in many ecological analyses and demographic models. In this demonstration you will

More information

(5) Multi-parameter models - Gibbs sampling. ST440/540: Applied Bayesian Analysis

(5) Multi-parameter models - Gibbs sampling. ST440/540: Applied Bayesian Analysis Summarizing a posterior Given the data and prior the posterior is determined Summarizing the posterior gives parameter estimates, intervals, and hypothesis tests Most of these computations are integrals

More information

A Tutorial on Learning with Bayesian Networks

A Tutorial on Learning with Bayesian Networks A utorial on Learning with Bayesian Networks David Heckerman Presented by: Krishna V Chengavalli April 21 2003 Outline Introduction Different Approaches Bayesian Networks Learning Probabilities and Structure

More information

Causal Inference with Big Data Sets

Causal Inference with Big Data Sets Causal Inference with Big Data Sets Marcelo Coca Perraillon University of Colorado AMC November 2016 1 / 1 Outlone Outline Big data Causal inference in economics and statistics Regression discontinuity

More information

Bayesian Inference on Joint Mixture Models for Survival-Longitudinal Data with Multiple Features. Yangxin Huang

Bayesian Inference on Joint Mixture Models for Survival-Longitudinal Data with Multiple Features. Yangxin Huang Bayesian Inference on Joint Mixture Models for Survival-Longitudinal Data with Multiple Features Yangxin Huang Department of Epidemiology and Biostatistics, COPH, USF, Tampa, FL yhuang@health.usf.edu January

More information

Causal Modeling in Environmental Epidemiology. Joel Schwartz Harvard University

Causal Modeling in Environmental Epidemiology. Joel Schwartz Harvard University Causal Modeling in Environmental Epidemiology Joel Schwartz Harvard University When I was Young What do I mean by Causal Modeling? What would have happened if the population had been exposed to a instead

More information

Comparative effectiveness of dynamic treatment regimes

Comparative effectiveness of dynamic treatment regimes Comparative effectiveness of dynamic treatment regimes An application of the parametric g- formula Miguel Hernán Departments of Epidemiology and Biostatistics Harvard School of Public Health www.hsph.harvard.edu/causal

More information

Multistate models and recurrent event models

Multistate models and recurrent event models Multistate models Multistate models and recurrent event models Patrick Breheny December 10 Patrick Breheny Survival Data Analysis (BIOS 7210) 1/22 Introduction Multistate models In this final lecture,

More information

Mixture modelling of recurrent event times with long-term survivors: Analysis of Hutterite birth intervals. John W. Mac McDonald & Alessandro Rosina

Mixture modelling of recurrent event times with long-term survivors: Analysis of Hutterite birth intervals. John W. Mac McDonald & Alessandro Rosina Mixture modelling of recurrent event times with long-term survivors: Analysis of Hutterite birth intervals John W. Mac McDonald & Alessandro Rosina Quantitative Methods in the Social Sciences Seminar -

More information

Sample Size Determination

Sample Size Determination Sample Size Determination 018 The number of subjects in a clinical study should always be large enough to provide a reliable answer to the question(s addressed. The sample size is usually determined by

More information

Whether to use MMRM as primary estimand.

Whether to use MMRM as primary estimand. Whether to use MMRM as primary estimand. James Roger London School of Hygiene & Tropical Medicine, London. PSI/EFSPI European Statistical Meeting on Estimands. Stevenage, UK: 28 September 2015. 1 / 38

More information

ANALYTIC COMPARISON. Pearl and Rubin CAUSAL FRAMEWORKS

ANALYTIC COMPARISON. Pearl and Rubin CAUSAL FRAMEWORKS ANALYTIC COMPARISON of Pearl and Rubin CAUSAL FRAMEWORKS Content Page Part I. General Considerations Chapter 1. What is the question? 16 Introduction 16 1. Randomization 17 1.1 An Example of Randomization

More information

Group Sequential Designs: Theory, Computation and Optimisation

Group Sequential Designs: Theory, Computation and Optimisation Group Sequential Designs: Theory, Computation and Optimisation Christopher Jennison Department of Mathematical Sciences, University of Bath, UK http://people.bath.ac.uk/mascj 8th International Conference

More information

Bayesian Networks Inference with Probabilistic Graphical Models

Bayesian Networks Inference with Probabilistic Graphical Models 4190.408 2016-Spring Bayesian Networks Inference with Probabilistic Graphical Models Byoung-Tak Zhang intelligence Lab Seoul National University 4190.408 Artificial (2016-Spring) 1 Machine Learning? Learning

More information

Analysis of Longitudinal Data. Patrick J. Heagerty PhD Department of Biostatistics University of Washington

Analysis of Longitudinal Data. Patrick J. Heagerty PhD Department of Biostatistics University of Washington Analysis of Longitudinal Data Patrick J Heagerty PhD Department of Biostatistics University of Washington Auckland 8 Session One Outline Examples of longitudinal data Scientific motivation Opportunities

More information

A Bayesian Machine Learning Approach for Optimizing Dynamic Treatment Regimes

A Bayesian Machine Learning Approach for Optimizing Dynamic Treatment Regimes A Bayesian Machine Learning Approach for Optimizing Dynamic Treatment Regimes Thomas A. Murray, (tamurray@mdanderson.org), Ying Yuan, (yyuan@mdanderson.org), and Peter F. Thall (rex@mdanderson.org) Department

More information

Analysing longitudinal data when the visit times are informative

Analysing longitudinal data when the visit times are informative Analysing longitudinal data when the visit times are informative Eleanor Pullenayegum, PhD Scientist, Hospital for Sick Children Associate Professor, University of Toronto eleanor.pullenayegum@sickkids.ca

More information

Lecture 13 Fundamentals of Bayesian Inference

Lecture 13 Fundamentals of Bayesian Inference Lecture 13 Fundamentals of Bayesian Inference Dennis Sun Stats 253 August 11, 2014 Outline of Lecture 1 Bayesian Models 2 Modeling Correlations Using Bayes 3 The Universal Algorithm 4 BUGS 5 Wrapping Up

More information

Causality II: How does causal inference fit into public health and what it is the role of statistics?

Causality II: How does causal inference fit into public health and what it is the role of statistics? Causality II: How does causal inference fit into public health and what it is the role of statistics? Statistics for Psychosocial Research II November 13, 2006 1 Outline Potential Outcomes / Counterfactual

More information

Approximation of Survival Function by Taylor Series for General Partly Interval Censored Data

Approximation of Survival Function by Taylor Series for General Partly Interval Censored Data Malaysian Journal of Mathematical Sciences 11(3): 33 315 (217) MALAYSIAN JOURNAL OF MATHEMATICAL SCIENCES Journal homepage: http://einspem.upm.edu.my/journal Approximation of Survival Function by Taylor

More information

Incorporating cost in Bayesian variable selection, with application to cost-effective measurement of quality of health care

Incorporating cost in Bayesian variable selection, with application to cost-effective measurement of quality of health care Incorporating cost in Bayesian variable selection, with application to cost-effective measurement of quality of health care Dimitris Fouskakis, Department of Mathematics, School of Applied Mathematical

More information

Estimation of Optimal Treatment Regimes Via Machine Learning. Marie Davidian

Estimation of Optimal Treatment Regimes Via Machine Learning. Marie Davidian Estimation of Optimal Treatment Regimes Via Machine Learning Marie Davidian Department of Statistics North Carolina State University Triangle Machine Learning Day April 3, 2018 1/28 Optimal DTRs Via ML

More information

Deep Temporal Generative Models of. Rahul Krishnan, Uri Shalit, David Sontag

Deep Temporal Generative Models of. Rahul Krishnan, Uri Shalit, David Sontag Deep Temporal Generative Models of Rahul Krishnan, Uri Shalit, David Sontag Patient timeline Jan 1 Feb 12 May 15 Blood pressure = 130 WBC count = 6*10 9 /L Temperature = 98 F A1c = 6.6% Precancerous cells

More information

Markov Chain Monte Carlo

Markov Chain Monte Carlo Markov Chain Monte Carlo Jamie Monogan University of Georgia Spring 2013 For more information, including R programs, properties of Markov chains, and Metropolis-Hastings, please see: http://monogan.myweb.uga.edu/teaching/statcomp/mcmc.pdf

More information

McGill University. Department of Epidemiology and Biostatistics. Bayesian Analysis for the Health Sciences. Course EPIB-675.

McGill University. Department of Epidemiology and Biostatistics. Bayesian Analysis for the Health Sciences. Course EPIB-675. McGill University Department of Epidemiology and Biostatistics Bayesian Analysis for the Health Sciences Course EPIB-675 Lawrence Joseph Bayesian Analysis for the Health Sciences EPIB-675 3 credits Instructor:

More information

Potential Outcomes and Causal Inference I

Potential Outcomes and Causal Inference I Potential Outcomes and Causal Inference I Jonathan Wand Polisci 350C Stanford University May 3, 2006 Example A: Get-out-the-Vote (GOTV) Question: Is it possible to increase the likelihood of an individuals

More information

Spatial discrete hazards using Hierarchical Bayesian Modeling

Spatial discrete hazards using Hierarchical Bayesian Modeling Spatial discrete hazards using Hierarchical Bayesian Modeling Mathias Graf ETH Zurich, Institute for Structural Engineering, Group Risk & Safety 1 Papers -Maes, M.A., Dann M., Sarkar S., and Midtgaard,

More information

Statistical models for estimating the effects of intermediate variables in the presence of time-dependent confounders

Statistical models for estimating the effects of intermediate variables in the presence of time-dependent confounders Statistical models for estimating the effects of intermediate variables in the presence of time-dependent confounders Dissertation zur Erlangung des Dotorgrades der Faultät für Mathemati und Physi der

More information

The lmm Package. May 9, Description Some improved procedures for linear mixed models

The lmm Package. May 9, Description Some improved procedures for linear mixed models The lmm Package May 9, 2005 Version 0.3-4 Date 2005-5-9 Title Linear mixed models Author Original by Joseph L. Schafer . Maintainer Jing hua Zhao Description Some improved

More information

Multistate models and recurrent event models

Multistate models and recurrent event models and recurrent event models Patrick Breheny December 6 Patrick Breheny University of Iowa Survival Data Analysis (BIOS:7210) 1 / 22 Introduction In this final lecture, we will briefly look at two other

More information

SEQUENTIAL MULTIPLE ASSIGNMENT RANDOMIZATION TRIALS WITH ENRICHMENT (SMARTER) DESIGN

SEQUENTIAL MULTIPLE ASSIGNMENT RANDOMIZATION TRIALS WITH ENRICHMENT (SMARTER) DESIGN SEQUENTIAL MULTIPLE ASSIGNMENT RANDOMIZATION TRIALS WITH ENRICHMENT (SMARTER) DESIGN Ying Liu Division of Biostatistics, Medical College of Wisconsin Yuanjia Wang Department of Biostatistics & Psychiatry,

More information

Metric Predicted Variable on One Group

Metric Predicted Variable on One Group Metric Predicted Variable on One Group Tim Frasier Copyright Tim Frasier This work is licensed under the Creative Commons Attribution 4.0 International license. Click here for more information. Prior Homework

More information

Bayesian Networks in Educational Assessment

Bayesian Networks in Educational Assessment Bayesian Networks in Educational Assessment Estimating Parameters with MCMC Bayesian Inference: Expanding Our Context Roy Levy Arizona State University Roy.Levy@asu.edu 2017 Roy Levy MCMC 1 MCMC 2 Posterior

More information

36-463/663Multilevel and Hierarchical Models

36-463/663Multilevel and Hierarchical Models 36-463/663Multilevel and Hierarchical Models From Bayes to MCMC to MLMs Brian Junker 132E Baker Hall brian@stat.cmu.edu 1 Outline Bayesian Statistics and MCMC Distribution of Skill Mastery in a Population

More information

Package lmm. R topics documented: March 19, Version 0.4. Date Title Linear mixed models. Author Joseph L. Schafer

Package lmm. R topics documented: March 19, Version 0.4. Date Title Linear mixed models. Author Joseph L. Schafer Package lmm March 19, 2012 Version 0.4 Date 2012-3-19 Title Linear mixed models Author Joseph L. Schafer Maintainer Jing hua Zhao Depends R (>= 2.0.0) Description Some

More information

January NCCTG Protocol No: N0927 Opened: September 3, 2009

January NCCTG Protocol No: N0927 Opened: September 3, 2009 January 2011 0915-1 RTOG Protocol No: 0915 Protocol Status: NCCTG Protocol No: N0927 Opened: September 3, 2009 Title: A Randomized Phase II Study Comparing 2 Stereotactic Body Radiation Therapy (SBRT)

More information

FAV i R This paper is produced mechanically as part of FAViR. See for more information.

FAV i R This paper is produced mechanically as part of FAViR. See  for more information. Bayesian Claim Severity Part 2 Mixed Exponentials with Trend, Censoring, and Truncation By Benedict Escoto FAV i R This paper is produced mechanically as part of FAViR. See http://www.favir.net for more

More information

Causal Effect Models for Realistic Individualized Treatment and Intention to Treat Rules

Causal Effect Models for Realistic Individualized Treatment and Intention to Treat Rules University of California, Berkeley From the SelectedWorks of Maya Petersen March, 2007 Causal Effect Models for Realistic Individualized Treatment and Intention to Treat Rules Mark J van der Laan, University

More information

Optimal Patient-specific Post-operative Surveillance for Vascular Surgeries

Optimal Patient-specific Post-operative Surveillance for Vascular Surgeries Optimal Patient-specific Post-operative Surveillance for Vascular Surgeries Peter I. Frazier, Shanshan Zhang, Pranav Hanagal School of Operations Research & Information Engineering, Cornell University

More information

Lecture 7 Time-dependent Covariates in Cox Regression

Lecture 7 Time-dependent Covariates in Cox Regression Lecture 7 Time-dependent Covariates in Cox Regression So far, we ve been considering the following Cox PH model: λ(t Z) = λ 0 (t) exp(β Z) = λ 0 (t) exp( β j Z j ) where β j is the parameter for the the

More information

Discussion on Fygenson (2007, Statistica Sinica): a DS Perspective

Discussion on Fygenson (2007, Statistica Sinica): a DS Perspective 1 Discussion on Fygenson (2007, Statistica Sinica): a DS Perspective Chuanhai Liu Purdue University 1. Introduction In statistical analysis, it is important to discuss both uncertainty due to model choice

More information

Targeted Learning for High-Dimensional Variable Importance

Targeted Learning for High-Dimensional Variable Importance Targeted Learning for High-Dimensional Variable Importance Alan Hubbard, Nima Hejazi, Wilson Cai, Anna Decker Division of Biostatistics University of California, Berkeley July 27, 2016 for Centre de Recherches

More information

Bayesian Meta-analysis with Hierarchical Modeling Brian P. Hobbs 1

Bayesian Meta-analysis with Hierarchical Modeling Brian P. Hobbs 1 Bayesian Meta-analysis with Hierarchical Modeling Brian P. Hobbs 1 Division of Biostatistics, School of Public Health, University of Minnesota, Mayo Mail Code 303, Minneapolis, Minnesota 55455 0392, U.S.A.

More information

Bayesian Inference for Regression Parameters

Bayesian Inference for Regression Parameters Bayesian Inference for Regression Parameters 1 Bayesian inference for simple linear regression parameters follows the usual pattern for all Bayesian analyses: 1. Form a prior distribution over all unknown

More information

Modelling geoadditive survival data

Modelling geoadditive survival data Modelling geoadditive survival data Thomas Kneib & Ludwig Fahrmeir Department of Statistics, Ludwig-Maximilians-University Munich 1. Leukemia survival data 2. Structured hazard regression 3. Mixed model

More information

Instrumental variables estimation in the Cox Proportional Hazard regression model

Instrumental variables estimation in the Cox Proportional Hazard regression model Instrumental variables estimation in the Cox Proportional Hazard regression model James O Malley, Ph.D. Department of Biomedical Data Science The Dartmouth Institute for Health Policy and Clinical Practice

More information

Treatment Effects. Christopher Taber. September 6, Department of Economics University of Wisconsin-Madison

Treatment Effects. Christopher Taber. September 6, Department of Economics University of Wisconsin-Madison Treatment Effects Christopher Taber Department of Economics University of Wisconsin-Madison September 6, 2017 Notation First a word on notation I like to use i subscripts on random variables to be clear

More information

Bayesian Estimation of the Time of a Linear Trend in Risk-Adjusted Control Charts

Bayesian Estimation of the Time of a Linear Trend in Risk-Adjusted Control Charts Bayesian Estimation of the Time of a Linear Trend in Risk-Adjusted Control Charts Hassan Assareh, Ian Smith, Kerrie Mengersen Abstract Change point detection is recognized as an essential tool of root

More information

Lecture 1 January 18

Lecture 1 January 18 STAT 263/363: Experimental Design Winter 2016/17 Lecture 1 January 18 Lecturer: Art B. Owen Scribe: Julie Zhu Overview Experiments are powerful because you can conclude causality from the results. In most

More information

An Introduction to WinBUGS PIMS Collaborative Research Group Summer School on Bayesian Modeling and Computation

An Introduction to WinBUGS PIMS Collaborative Research Group Summer School on Bayesian Modeling and Computation An Introduction to PIMS Collaborative Research Group Summer School on Bayesian Modeling and Computation 14 th July, 2008 Sebastien Haneuse, PhD Group Health Center for Health Studies Department of Biostatistics,

More information

Bayesian regression tree models for causal inference: regularization, confounding and heterogeneity

Bayesian regression tree models for causal inference: regularization, confounding and heterogeneity Bayesian regression tree models for causal inference: regularization, confounding and heterogeneity P. Richard Hahn, Jared Murray, and Carlos Carvalho June 22, 2017 The problem setting We want to estimate

More information

BNP survival regression with variable dimension covariate vector

BNP survival regression with variable dimension covariate vector BNP survival regression with variable dimension covariate vector Peter Müller, UT Austin Survival 0.0 0.2 0.4 0.6 0.8 1.0 Truth Estimate Survival 0.0 0.2 0.4 0.6 0.8 1.0 Truth Estimate BC, BRAF TT (left),

More information

Chapter 11. Correlation and Regression

Chapter 11. Correlation and Regression Chapter 11. Correlation and Regression The word correlation is used in everyday life to denote some form of association. We might say that we have noticed a correlation between foggy days and attacks of

More information

McGill University. Department of Epidemiology and Biostatistics. Bayesian Analysis for the Health Sciences. Course EPIB-682.

McGill University. Department of Epidemiology and Biostatistics. Bayesian Analysis for the Health Sciences. Course EPIB-682. McGill University Department of Epidemiology and Biostatistics Bayesian Analysis for the Health Sciences Course EPIB-682 Lawrence Joseph Intro to Bayesian Analysis for the Health Sciences EPIB-682 2 credits

More information

Bayesian inference. Fredrik Ronquist and Peter Beerli. October 3, 2007

Bayesian inference. Fredrik Ronquist and Peter Beerli. October 3, 2007 Bayesian inference Fredrik Ronquist and Peter Beerli October 3, 2007 1 Introduction The last few decades has seen a growing interest in Bayesian inference, an alternative approach to statistical inference.

More information

Measuring Social Influence Without Bias

Measuring Social Influence Without Bias Measuring Social Influence Without Bias Annie Franco Bobbie NJ Macdonald December 9, 2015 The Problem CS224W: Final Paper How well can statistical models disentangle the effects of social influence from

More information

Metric Predicted Variable With One Nominal Predictor Variable

Metric Predicted Variable With One Nominal Predictor Variable Metric Predicted Variable With One Nominal Predictor Variable Tim Frasier Copyright Tim Frasier This work is licensed under the Creative Commons Attribution 4.0 International license. Click here for more

More information

Cluster Analysis using SaTScan

Cluster Analysis using SaTScan Cluster Analysis using SaTScan Summary 1. Statistical methods for spatial epidemiology 2. Cluster Detection What is a cluster? Few issues 3. Spatial and spatio-temporal Scan Statistic Methods Probability

More information

arxiv: v5 [stat.me] 13 Feb 2018

arxiv: v5 [stat.me] 13 Feb 2018 arxiv: arxiv:1602.07933 BOOTSTRAP INFERENCE WHEN USING MULTIPLE IMPUTATION By Michael Schomaker and Christian Heumann University of Cape Town and Ludwig-Maximilians Universität München arxiv:1602.07933v5

More information

Multivariate Survival Analysis

Multivariate Survival Analysis Multivariate Survival Analysis Previously we have assumed that either (X i, δ i ) or (X i, δ i, Z i ), i = 1,..., n, are i.i.d.. This may not always be the case. Multivariate survival data can arise in

More information

Lectures of STA 231: Biostatistics

Lectures of STA 231: Biostatistics Lectures of STA 231: Biostatistics Second Semester Academic Year 2016/2017 Text Book Biostatistics: Basic Concepts and Methodology for the Health Sciences (10 th Edition, 2014) By Wayne W. Daniel Prepared

More information