On ARL-unbiased c-charts for i.i.d. and INAR(1) Poisson counts

Similar documents
ARL-unbiased geometric control charts for high-yield processes

On ARL-unbiased c charts for i.i.d. and INAR(1) Poisson counts

A Modified Poisson Exponentially Weighted Moving Average Chart Based on Improved Square Root Transformation

ON HITTING TIMES FOR MARKOV TIME SERIES OF COUNTS WITH APPLICATIONS TO QUALITY CONTROL

THE DETECTION OF SHIFTS IN AUTOCORRELATED PROCESSES WITH MR AND EWMA CHARTS

A Theoretically Appropriate Poisson Process Monitor

Control charting normal variance reflections, curiosities, and recommendations

Sequential Procedure for Testing Hypothesis about Mean of Latent Gaussian Process

CUMULATIVE SUM CHARTS FOR HIGH YIELD PROCESSES

Control charts are used for monitoring the performance of a quality characteristic. They assist process

CONTROL CHARTS FOR THE GENERALIZED POISSON PROCESS WITH UNDER-DISPERSION

On detection of unit roots generalizing the classic Dickey-Fuller approach

Statistical quality control (SQC)

Jumps in binomial AR(1) processes

On Monitoring Shift in the Mean Processes with. Vector Autoregressive Residual Control Charts of. Individual Observation

Control Charts for Monitoring the Zero-Inflated Generalized Poisson Processes

An Economic Alternative to the c Chart

Statistics Ph.D. Qualifying Exam: Part II November 9, 2002

Formulas and Tables by Mario F. Triola

Directionally Sensitive Multivariate Statistical Process Control Methods

Statistical Process Control SCM Pearson Education, Inc. publishing as Prentice Hall

Change Point Estimation of the Process Fraction Non-conforming with a Linear Trend in Statistical Process Control

An Adaptive Exponentially Weighted Moving Average Control Chart for Monitoring Process Variances

A New Demerit Control Chart for Monitoring the Quality of Multivariate Poisson Processes. By Jeh-Nan Pan Chung-I Li Min-Hung Huang

Quality Control & Statistical Process Control (SPC)

Control of Manufacturing Processes

Approximating the step change point of the process fraction nonconforming using genetic algorithm to optimize the likelihood function

Binomial and Poisson Probability Distributions

Robust control charts for time series data

ON CONSTRUCTING T CONTROL CHARTS FOR RETROSPECTIVE EXAMINATION. Gunabushanam Nedumaran Oracle Corporation 1133 Esters Road #602 Irving, TX 75061

The Robustness of the Multivariate EWMA Control Chart

Lecture 3. Discrete Random Variables

HANDBOOK OF APPLICABLE MATHEMATICS

Design and Implementation of CUSUM Exceedance Control Charts for Unknown Location

STATISTICAL PROCESS CONTROL

The Effect of Level of Significance (α) on the Performance of Hotelling-T 2 Control Chart

Optimal SPRT and CUSUM Procedures using Compressed Limit Gauges

Chapter 1 Statistical Inference

Formulas and Tables. for Elementary Statistics, Tenth Edition, by Mario F. Triola Copyright 2006 Pearson Education, Inc. ˆp E p ˆp E Proportion

Statistical Process Control for Multivariate Categorical Processes

Zero-Inflated Models in Statistical Process Control

Deccan Education Society s FERGUSSON COLLEGE, PUNE (AUTONOMOUS) SYLLABUS UNDER AUTOMONY. SECOND YEAR B.Sc. SEMESTER - III

2.830J / 6.780J / ESD.63J Control of Manufacturing Processes (SMA 6303) Spring 2008

Chapter 2 Detection of Changes in INAR Models

A Control Chart for Time Truncated Life Tests Using Exponentiated Half Logistic Distribution

MONITORING BIVARIATE PROCESSES WITH A SYNTHETIC CONTROL CHART BASED ON SAMPLE RANGES

Variations in a manufacturing process can be categorized into common cause and special cause variations. In the presence of

On the Distribution of Hotelling s T 2 Statistic Based on the Successive Differences Covariance Matrix Estimator

Because of the global market that considers customer satisfaction as a primary objective, quality has been established as a key

Early Detection of a Change in Poisson Rate After Accounting For Population Size Effects

CHAPTER-II Control Charts for Fraction Nonconforming using m-of-m Runs Rules

Optimal Cusum Control Chart for Censored Reliability Data with Log-logistic Distribution

ON THE CONSEQUENCES OF MISSPECIFING ASSUMPTIONS CONCERNING RESIDUALS DISTRIBUTION IN A REPEATED MEASURES AND NONLINEAR MIXED MODELLING CONTEXT

Normalizing the I Control Chart

Poisson INAR processes with serial and seasonal correlation

Learning Objectives for Stat 225

CSE 312 Final Review: Section AA

Practice Problems Section Problems

Chapter 5. Chapter 5 sections

3 Joint Distributions 71

Concentration of Measures by Bounded Couplings

SYSM 6303: Quantitative Introduction to Risk and Uncertainty in Business Lecture 4: Fitting Data to Distributions

December 19, Probability Theory Instituto Superior Técnico. Poisson Convergence. João Brazuna. Weak Law of Small Numbers

Run sum control charts for the monitoring of process variability

Stat 315: HW #6. Fall Due: Wednesday, October 10, 2018

Expectation. DS GA 1002 Statistical and Mathematical Models. Carlos Fernandez-Granda

Testing Statistical Hypotheses

Faculty of Science and Technology MASTER S THESIS

Surveillance of Infectious Disease Data using Cumulative Sum Methods

On the performance of Shewhart-type synthetic and runs-rules charts combined with an chart

Confirmation Sample Control Charts

Lecture 13. Poisson Distribution. Text: A Course in Probability by Weiss 5.5. STAT 225 Introduction to Probability Models February 16, 2014

Chapter 10: Statistical Quality Control

Statistics 910, #5 1. Regression Methods

Quality Control The ASTA team

The exact bias of S 2 in linear panel regressions with spatial autocorrelation SFB 823. Discussion Paper. Christoph Hanck, Walter Krämer

Optimal global rates of convergence for interpolation problems with random design

Cumulative probability control charts for geometric and exponential process characteristics

Section II: Assessing Chart Performance. (Jim Benneyan)

One-Sample Numerical Data

Conditional Maximum Likelihood Estimation of the First-Order Spatial Non-Negative Integer-Valued Autoregressive (SINAR(1,1)) Model

Econometrics. Week 4. Fall Institute of Economic Studies Faculty of Social Sciences Charles University in Prague

Modeling and Performance Analysis with Discrete-Event Simulation

DA Freedman Notes on the MLE Fall 2003

Computer Science, Informatik 4 Communication and Distributed Systems. Simulation. Discrete-Event System Simulation. Dr.

IMPROVING TWO RESULTS IN MULTIPLE TESTING

A Multivariate Process Variability Monitoring Based on Individual Observations

Module B1: Multivariate Process Control

Basic Concepts of Inference

Introduction to Statistical Data Analysis Lecture 3: Probability Distributions

SCIENCE & TECHNOLOGY

A Power Analysis of Variable Deletion Within the MEWMA Control Chart Statistic

1 Inverse Transform Method and some alternative algorithms

Performance of Conventional X-bar Chart for Autocorrelated Data Using Smaller Sample Sizes

Large Sample Properties of Estimators in the Classical Linear Regression Model

Songklanakarin Journal of Science and Technology SJST R1 Sukparungsee

CONTROL CHARTS FOR MULTIVARIATE NONLINEAR TIME SERIES

Asymptotic Statistics-III. Changliang Zou

Evaluation of X and S charts when Standards Vary Randomly *Aamir Saghir Department of Statistics, UAJK Muzaffrabad, Pakistan.

A new multivariate CUSUM chart using principal components with a revision of Crosier's chart

Transcription:

On ARL-unbiased c-charts for iid and INAR(1) Poisson counts Manuel Cabral Morais (1) with Sofia Paulino (2) and Sven Knoth (3) (1) Department of Mathematics & CEMAT IST, ULisboa, Portugal (2) IST, ULisboa, Portugal (3) Department of Economics and Social Sciences Helmut Schmidt University, Germany IST, ULisboa Lisbon, February 17, 2016

Introduction The iid case The INAR(1) case Final thoughts Agenda 1 Introduction Initial thoughts and motivation 2 The iid case Revisiting the c-chart Quantile based control limits ARL-unbiased c charts for λ 3 The INAR(1) case Definitions and properties ARL-unbiased c chart for λ/(1 β) 4 Final thoughts On ARL-unbiased c-charts for iid and INAR(1) Poisson counts

Introduction The iid case The INAR(1) case Final thoughts Initial thoughts and motivation Processes of counts of nonconformities arise frequently in SPC Their marginal distribution is usually assumed to be Poisson We often deal with autocorrelated count processes but falsely assume serial independence, namely while planning a control chart to monitor the process mean On ARL-unbiased c-charts for iid and INAR(1) Poisson counts

Introduction The iid case The INAR(1) case Final thoughts Initial thoughts and motivation Dealing with a nonnegative, discrete and asymmetrical (eg right-skewed) distribution (eg Poisson) can prevent us to: set a pre-specified in-control ARL; deal with a positive lower control limit; be able to quickly detect small and moderate decreases in the process mean; define an ARL-unbiased control chart (Pignatiello et al, 1995; Acosta-Mejía, 1999) in the sense that it takes longer, in average, to trigger a false alarm than to detect any shifts in the process mean On ARL-unbiased c-charts for iid and INAR(1) Poisson counts

Introduction The iid case The INAR(1) case Final thoughts Revisiting the c-chart The c chart is one of the most popular procedures to control the expected number of defects (λ) in a sample of n units Control statistic: total number of defects in the t th sample, X t Distribution: X t indep X Poisson(λ), t N Target mean: λ 0 Process mean: λ = λ 0 + δ 3 σ control limits: LCL = max {0, λ 0 3 } λ 0 (δ is the magnitude of the shift) UCL = λ 0 + 3 λ 0 Trigger a signal and deem the process out-of-control at sample t if X t [LCL, UCL] Performance measure run length: RL(δ) Geometric( ξ(δ) = P[X [LCL, UCL] δ] ) ARL(δ) = 1/ξ(δ) On ARL-unbiased c-charts for iid and INAR(1) Poisson counts

Introduction The iid case The INAR(1) case Final thoughts Revisiting the c-chart If λ 0 9 then LCL = 0 and the chart triggers false alarms more frequently than valid signals in the presence of any decrease in λ For λ 0 > 9, the ARL function of a c chart with 3 σ control limits attains its maximum value at» 1 δ UCL! UCL LCL+1 (λ 0) = argmax δ ( λ0,+ )ARL(δ) = λ0 < 0 (LCL 1)! argmax 30 25 20 15 10 05 00 20 40 60 80 100 λ 0 On ARL-unbiased c-charts for iid and INAR(1) Poisson counts

Introduction The iid case The INAR(1) case Final thoughts Revisiting the c-chart Some variants of the c chart c charts with 3 σ control limits may behave poorly when it comes to the detection of decreases in λ Alternative approaches (Aebtarm & Bouguila, 2011): transforming data; standardizing data; optimizing control limits Best overall c chart (optimal control limits) Proposed by Ryan & Schwertman (1997), as defined by Aebtarm & Bouguila (2011) control limits are obtained by linear regression based on a table of the best c chart limits for several values of λ 0: LCL = 15307 + 10212λ 0 32197 λ 0 ; UCL = 06182 + 09996λ 0 + 30303 λ 0 On ARL-unbiased c-charts for iid and INAR(1) Poisson counts

Introduction The iid case The INAR(1) case Final thoughts Quantile based control limits Control limits based on quantiles can be an appropriate alternative to the 3 σ limits, but the discrete character of the Poisson distribution makes it difficult to achieve the pre-specified probability of false alarm α = 00027 Quantile based control limits Requiring that P(X [LCL, UCL] δ = 0) α and setting α = α LCL + α UCL leads to obtention of LCL and UCL such that: P(X < LCL δ = 0) α LCL ; P(X > UCL δ = 0) α UCL Thus, the quantile based LCL (resp UCL) is the largest (resp smallest) nonneg integer satisfying the 1st (resp 2nd) condition Performance measure: 1 ARL(δ) = ξ(δ) = P(X [LCL, UCL] δ) ; ARL(0) α 1 On ARL-unbiased c-charts for iid and INAR(1) Poisson counts

Introduction The iid case The INAR(1) case Final thoughts 2 2015-09-30-ARL-unbiased-c-chart-Table1Figure1nb Quantile based control limits Example 1: Comparing ARL performance (λ 0 = 8) Out[525]= 518209 Out[529]= 1 Unlike the 3 σ limits, the quantile based limits lead to a c-chart with an ARL function not as biased as the one of the R&S c-chart Out[530]= 18 Out[531]= 101437 ARL( ) 10 6 10 5 10 4 Out[532]= 1000 100 10 Figure: ARL(δ) for: 3 σ control limits, [0, 16] (dotted lines); R&S control limits, [1, 17] (dashed lines); quantile based control limits, [1, 18] (thin solid lines; α = 00027, α LCL = α UCL ) -4-2 0 2 4 On ARL-unbiased c-charts for iid and INAR(1) Poisson counts

Introduction The iid case The INAR(1) case Final thoughts ARL-unbiased c charts for λ Inspired by UMPU (Uniformly Most Powerful Unbiased) tests (Lehmann, 1986, pp 134-140) and randomized tests (rarely used in SPC when dealing with discrete distributions), we defined a c chart, with quantile based control limits, that triggers a signal with: probability one if the sample number of defects is below LCL or above UCL; probabilities γ LCL and γ UCL if the sample number of defects is equal to LCL and UCL, resp We are basically considering H 0 : λ = λ 0 vs H 1 : λ λ 0 1 if x < LCL or x > UCL γ LCL if x = LCL φ(x) = P(Reject H 0 X = x) = γ UCL if x = UCL 0 if LCL < x < UCL On ARL-unbiased c-charts for iid and INAR(1) Poisson counts

Introduction The iid case The INAR(1) case Final thoughts ARL-unbiased c charts for λ The randomization probabilities satisfy { E λ0 [φ(x )] = α (prob of a false alarm = α) (γ LCL, γ UCL ) : E λ0 [X φ(x )] = α E λ0 (X ) (unbiased ARL) The solution of this system of linear equations: γ LCL = where a = P λ0 (LCL), de bf ad bc c = LCL P λ0 (LCL), and b = P λ0 (UCL) e = α 1 + UCL x=lclp λ0 (x) γ LCL = d = UCL P λ0 (UCL) f = (α 1) E λ0 (X ) + UCL x=lclx P λ0 (x) af ce ad bc, On ARL-unbiased c-charts for iid and INAR(1) Poisson counts

Introduction The iid case The INAR(1) case Final thoughts ARL-unbiased c charts for λ In order to rule out pairs of control limits leading to (γ LCL, γ UCL ) (0, 1) 2, the potentially useful (LCL, UCL) are restricted to the following grid of non-negative integer numbers: {(LCL, UCL) : L min LCL L max, U min UCL U max } The search for admissible values for (γ LCL, γ UCL ) starts with (LCL, UCL) = (L min, U min ) and stops as soon as an admissible solution is found L min = n max F 1 (max{0, F (U min 1) 1 + α}), G 1 o (max{0, G(U min 1) 1 + α}) n o 1 1 L max = min F (α), G (α) n U min = max F 1 (1 α), G 1 o (1 α) n 1 U max = min F (min{1, F (Lmax ) + 1 α}), G 1 o (min{1, G(L max ) + 1 α}) F (x) = P λ0 (X x) G(x) = 1 X x i =0 λ ξ Pp 0 (X = i) 0 F 1 (α) = min{x N 0 : F (x) α} F 1 (α) = min{x N 0 : F (x) > α} G 1 (α) = min{x N 0 : G(x) α} G 1 (α) = min{x N 0 : G(x) > α} On ARL-unbiased c-charts for iid and INAR(1) Poisson counts

Introduction General::stop : Further output The of iid LessEqual::nord case will be suppressed The during INAR(1) thiscase calculation Final thoughts ARL-unbiased GreaterEqual::nord c charts for: λinvalid comparison with ComplexInfinity attempted General::stop : Further output of GreaterEqual::nord will be suppressed during this calculation Example 2: Comparing ARL performance (λ 0 = 8) When 0=8 and =00027 we get LCLunbiased=1, UCLunbiased=18, L= 0482414, U=0444451, in-control ARL=37037, relative bias=915732 10-11 1000 100 10 ARL( ) -4-2 0 2 4 Figure: ARL(δ) for: c chart with quantile based control limits (thin solid lines); ARL-unbiased c chart with [LCL, UCL] = [1, 18] & randomization prob (γ LCL, γ UCL ) = (0482414, 0444451) (α = 00027; thick solid lines) Randomization increases the prob of triggering a signal, thus smaller ARL values, yet a reasonable in-control ARL (1/00027 3704) On ARL-unbiased c-charts for iid and INAR(1) Poisson counts

Introduction The iid case The INAR(1) case Final thoughts ARL-unbiased c charts for λ ARL-unbiased designs Table: Limits of the search grid, quantile based control limits and randomization prob of the ARL-unbiased c charts, for λ 0 = 005, 01(01)1, 2 20, α = 00027 λ 0 L min L max LCL U min U max UCL γ LCL γ UCL 005 0 0 0 2 2 2 0002778 0031347 01 0 0 0 3 3 3 0002886 0562609 09 0 0 0 5 5 5 0005639 0032019 1 0 0 0 6 6 6 0006159 0686904 8 1 1 1 18 18 18 0482414 0444451 20 4 9 8 34 37 35 0566150 0549842 On ARL-unbiased c-charts for iid and INAR(1) Poisson counts

Introduction The iid case The INAR(1) case Final thoughts Time series of counts Arise in areas such as industry the monthly number of accidents in a manufacturing plant and the number of defects per sample have to be controlled When the time series consists only of small integer numbers, ARMA processes are of limited use, namely because the multiplication of an integer-valued rv by a real constant may lead to a non-integer rv A possible way out is to replace the scalar multiplication by a random operation, such as the binomial thinning operation On ARL-unbiased c-charts for iid and INAR(1) Poisson counts

Introduction The iid case The INAR(1) case Final thoughts Definitions and properties Binomial thinning operation Let X be a discrete rv with range N 0 and β a scalar in (0, 1) Then the binomial thinning operation on X results in another rv defined as follows: X β X = Y i, i=1 where {Y i : i N} is a sequence of iid Bernoulli(β) rv, independent of X β X arises from X by binomial thinning On ARL-unbiased c-charts for iid and INAR(1) Poisson counts

Introduction The iid case The INAR(1) case Final thoughts Definitions and properties INAR(1) process {X t : t Z} is said to be a first-order integer-valued autoregressive process if X t = β X t 1 + ɛ t, where: β (0, 1); represents the binomial thinning operator; {ɛ t : t Z} be a sequence of nonnegative integer-valued iid rv, with mean µ ɛ and variance σ 2 ɛ ; ɛ t and X t 1 are assumed to be independent rv; all thinning operations are performed independently of each other and of {ɛ t : t Z}; the thinning operations at time t are independent of {, X t 2, X t 1 } On ARL-unbiased c-charts for iid and INAR(1) Poisson counts

Introduction The iid case The INAR(1) case Final thoughts Definitions and properties Poisson INAR(1) process If ɛ t iid Poisson(λ), t Z, then {X t = β X t 1 + ɛ t : t Z} is said to be a Poisson INAR(1) process It is a second order weakly stationary process such that X t Poisson (λ/(1 β)), t Z Its transition probability matrix, P = [p i j ] i,j = [P(X t = j X t 1 = i)] i,j, has entries given by p i j p i j (λ, β) ix = P(β X t 1 = m X t 1 = i) P(ɛ t = j m) = m=0 min{i,j} X m=0! i β m (1 β) i m e λ λ j m, i, j N0 m (j m)! On ARL-unbiased c-charts for iid and INAR(1) Poisson counts

Introduction The iid case The INAR(1) case Final thoughts Definitions and properties c-chart for the mean of a Poisson INAR(1) process Control statistic: X t, t N Target value: λ 0 /(1 β 0 ) k σ control limits: LCL = max { 0, λ 0 /(1 β 0 ) k } λ 0 /(1 β 0 ) UCL = λ 0 /(1 β 0 ) + k λ 0 /(1 β 0 ), where k is a positive constant (Weiss, 2009, p 419) Parameters: λ = λ 0 + δ λ or β = β 0 + δ β, where δ λ and δ β are the magnitude of shifts in λ and β On ARL-unbiased c-charts for iid and INAR(1) Poisson counts

Introduction The iid case The INAR(1) case Final thoughts Definitions and properties Run length Let: y = LCL and x = UCL ; Then RL u (λ, β) = min{t : X t < y or X t > x X 0 = u} be the RL of the c-chart, conditional on X 0 = u (u {x,, y}), λ and β ARL u (λ, β) = e u [I Q(λ, β)] 1 1, (1) where e u : (u y + 1) th vector of the orthogonal basis for R (x y+1), Q(λ, β) = [p lk (λ, β)] x l,k=y, I: identity matrix with rank (x y + 1), 1: column-vector with (x y + 1) ones In addition, overall ARL function (Weiss and Testik, 2009): x ARL(λ, β) = 1 + ARL u (λ, β) P[X 1 (λ, β) = u] ARL-biased chart! u=y On ARL-unbiased c-charts for iid and INAR(1) Poisson counts

Introduction The iid case The INAR(1) case Final thoughts ARL-unbiased c chart for λ/(1 β) The ARL-unbiased c chart for the mean of a Poisson INAR(1) process triggers a signal at sample t with: probability one if the count of nonconformities, X t, is beyond the control limits L and U; probability γ L (resp γ U ) if X t is equal to L (resp U) Since the control statistics are dependent rv, we can no longer: explicitly define a search grid for the non-negative integers L and U in the terms of the false alarm rate; obtain (γ L, γ U ) by solving a system of linear equations On ARL-unbiased c-charts for iid and INAR(1) Poisson counts

Introduction The iid case The INAR(1) case Final thoughts ARL-unbiased c chart for λ/(1 β) RL of the ARL-unbiased c chart Randomizing the emission of a signal means considering Q(λ, β, γ L, γ U ) equal to 2 3 (1 γ L ) p L L p L L+1 p L U 1 (1 γ U ) p L U (1 γ L ) p L+1 L p L+1 L+1 p L+1 U 1 (1 γ U ) p L+1 U 6 7 4 (1 γ L ) p U 1 L p U 1 L+1 p U 1 U 1 (1 γ U ) p U 1 U 5 (1 γ L ) p U L p U L+1 p U U 1 (1 γ U ) p U U The associated overall ARL, ARL(λ, β, γ L, γ U ): 1 + (1 γ L ) ARL L (λ, β, γ L, γ U ) P[X 1(λ, β) = L] + X U 1 u=l+1arl u (λ, β, γ L, γ U ) P[X 1(λ, β) = u] + (1 γ U ) ARL U (λ, β, γ L, γ U ) P[X 1(λ, β) = U], where ARL u (λ, β, γ L, γ U ) is obtained from (1), by taking y = L, x = U and Q(λ, β, γ L, γ U ) instead of Q(λ, β) The procedure to obtain L, U, γ L, and γ U involves a nested secant rule, etc On ARL-unbiased c-charts for iid and INAR(1) Poisson counts

Introduction The iid case The INAR(1) case Final thoughts ARL-unbiased c chart for λ/(1 β) Example 3: Comparing ARL performance (λ 0 = 3, β 0 = 05) µ ARL(λ,β 0 ) 0 200 400 600 4 5 6 7 8 20 25 30 35 40 Figure: ARL(λ, β 0 ) for: c chart with 3 σ control limits (thin solid lines); ARL-unbiased c chart, [LCL, UCL] = [0, 15] & (γ L, γ U ) = (0696553, 0693760) (ARL = 3704; thick solid lines) λ On ARL-unbiased c-charts for iid and INAR(1) Poisson counts

Introduction The iid case The INAR(1) case Final thoughts ARL-unbiased c chart for λ/(1 β) Example 3 (cont d) µ ARL(λ 0,β) 0 200 400 600 3 375 5 75 15 Inf 00 02 04 06 08 10 β Figure: ARL(λ 0, β) for: c chart with 3 σ control limits (thin solid lines); ARL-unbiased c chart, [LCL, UCL] = [0, 15] & (γ L, γ U ) = (0681815, 0736829) (ARL = 3704; thick solid lines) On ARL-unbiased c-charts for iid and INAR(1) Poisson counts

Introduction The iid case The INAR(1) case Final thoughts Iid/INAR(1) Poisson output The ARL-unbiased c charts we derived have in-control ARL equal to the pre-specified and desired value, requires, in average, less time to trigger a signal in the presence of all shifts than to trigger a false alarm, tackle the curse of the null lower control limit and detect decreases in a timely fashion, in contrast to the c charts with 3 σ control limits On-going and future work Morais (2016a) (resp Morais, 2016b) has already derived an ARL-unbiased np (resp geometric) chart and improved what Acosta-Mejía (1999) (resp Zhang et al, 2004) termed as a nearly ARL-unbiased p (resp geometric) chart Derive an ARL-unbiased version of the CUSUM charts for both iid and INAR(1) output A few difficulties will arise with Poisson INAR(1) output On ARL-unbiased c-charts for iid and INAR(1) Poisson counts

Introduction The iid case The INAR(1) case Final thoughts Acosta-Mejía, CA (1999) Improved p-charts to monitor process quality IIE Transactions 31, 509 516 Aebtarm, S and Bouguila, N (2011) An empirical evaluation of attribute control charts for monitoring defects Expert Systems with Applications 38, 7869 7880 Knoth, S and Morais, MC (2013) On ARL-unbiased control charts In S Knoth, W Schmid, Frontiers in Statistical Quality Control 11, pp 95 116 Springer Lehmann, EL (1986) Testing Statistical Hypotheses (2nd edition) Pacific Grove, California: Wadsworth & Brooks/Cole Advanced Books & Software Montgomery, DC (2009) Introduction to Statistical Quality Control (6thedition) New York: John Wiley & Sons, Inc Morais, MC (2012) Real- and Integer-valued Time Series and Quality Control Charts Unpublished report Morais, MC (2016a) An ARL-unbiased np-chart Accepted for publication in Economic Quality Control volume 31, issue 1 (June 2016) Morais, MC (2016b) ARL-unbiased geometric and CCC G control charts In preparation Paulino, S, Morais, MC, Knoth, S (2016) An ARL-unbiased c-chart Accepted for publication in Quality and Reliability Engineering International Paulino, S, Morais, MC, Knoth, S (2015) On ARL-unbiased c-charts for INAR(1) Poisson counts Submitted for publication Pignatiello, J J, Jr, Acosta-Mejía, C A, Rao, BV (1995) The performance of control charts for monitoring process dispersion 4th Industrial Engineering Research Conference, 320 328 Ryan, TP and Schwertman, NC (1997) Optimal limits for attribute control charts Journal of Quality Technology 29, 86 98 Weiss, CH (2007) Controlling correlated processes of Poisson counts Quality and Reliability Engineering International 23, 741 754 Weiss, CH (2009) Categorical Time Series Analysis and Applications in Statistical Quality Control PhD Thesis, Fakultät fur Mathematik und Informatik der Universität Würzburg dissertationde Verlag im Internet GmbH Weiss, CH and Testik, MC (2009) CUSUM monitoring of first-order integer-valued autoregressive processes of Poisson counts Journal of Quality Technology 41, 389 400 Zhang, L, Govindaraju, K, Bebbington, M and Lai, CD (2004) On the statistical design of geometric control charts Quality Technology & Quantitative Management 2, 233 243 On ARL-unbiased c-charts for iid and INAR(1) Poisson counts