Efficiency Comparison Between Mean and Log-rank Tests for. Recurrent Event Time Data

Size: px
Start display at page:

Download "Efficiency Comparison Between Mean and Log-rank Tests for. Recurrent Event Time Data"

Transcription

1 Efficiency Comparison Between Mean and Log-rank Tests for Recurrent Event Time Data Wenbin Lu Department of Statistics, North Carolina State University, Raleigh, NC Summary. Recurrent event time data are common in biomedical follow-up studies, in which a study subject may experience repeated occurrences of an event of interest. In this paper, we evaluate two popular nonparametric tests for recurrent event time data in terms of their relative efficiency. One is the log-rank test for classical survival data and the other a more recently developed nonparametric test based on comparing mean recurrent rates. We show analytically that, somewhat surprisingly, the log-rank test that only makes use of time to the first occurrence could be more efficient than the test for mean occurrence rates that makes use of all available recurrence times, provided that subject-to-subject variation of recurrence times is large. Explicit formula are derived for asymptotic relative efficiencies under the frailty model. The findings are demonstrated via extensive simulations. Key words: Asymptotic relative efficiency, frailty model, log-rank test, proportional mean test, recurrent events, robust variance estimation.. Introduction The log-rank test (Mantel and Haenszel, 959) is perhaps the most widely used method in two-sample comparisons of treatment efficacies for time-to-event data. It is simple to use, nonparametric in nature and highly efficient under suitable assumptions and incorporates the

2 usual right censorship without complication. It is also closely related to the Cox proportional hazards regression model (Cox, 972). In fact, it is the score test of the partial likelihood (Cox, 975) under the Cox model assumption. For historical reasons in the development of rank tests and for efficiency considerations under nonproportional hazards, weighted log-rank tests, notably the Gehan, the Peto-Prentice and the G ρ -family have also received a great deal of attention; cf. Gehan (965), Peto and Peto (972), Prentice (978), Harrington and Fleming (982) and Fleming and Harrington (99). In many biomedical follow-up studies, as well as studies in other disciplines such as economics, sociology and software engineering, subjects often experience repeated occurrences of the event of interest, i.e. recurrence of the same type of event. Examples include repeated infections of certain diseases, attacks of asthma and epileptic seizures, among others. In fact, the present investigation was motivated by studies sponsored by the R.W. Johnson Pharmaceutical Research Institute on evaluation of treatments for epileptic seizures, which are known to have big subject-to-subject variation in terms of epileptic seizure counts. A major development in the analysis of recurrent event time data is due to Andersen and Gill (982), who introduced a multiplicative intensity model for multivariate counting process which mimics the Cox proportional hazards model for failure time data. Under their model assumption, the method of partial likelihood can be used to obtain semiparametrically efficient estimates of the regression parameters. However, due to within-subject dependency, the requirement for a multiplicative intensity is likely to be too stringent without including complicated time-dependent covariate adjustment. To avoid modelling the intensity, Pepe and Cai (993) proposed use of rate functions for recurrent event time data so that the regression relationship is through the rate function instead of the intensity. Further studies along the line can be found in Lawless and Nadeau (995) and Lawless, Nadeau and Cook (997), who also 2

3 suggested modelling the regression through the mean functions of counting processes. For a comprehensive discussion of the rate and mean function models, the asymptotic theory thereof, as well as their relationship to the Andersen-Gill multiplicative intensity model, we refer to Lin, Wei, Yang and Ying (2). Additional approaches to dealing with recurrent event time data can be found in Wang and Chang (999) and Chang and Wang (999). An important ingredient in the mean/rate function approach to recurrent event time data is that the simple score function derived under the Andersen-Gill multiplicative intensity model assumption is unbiased, at least asymptotically, thereby can be used for hypothesis testing and parameter estimation, provided the robust variance estimation is adopted. In the case of the two-sample problem of testing treatment difference, the resulting test is nonparametric in the sense that its validity does not require any parametric or semiparametric model assumption. Alternatively, the log-rank test can also be used for testing treatment difference when only time to first event is used and subsequent event times are ignored. At a first glance, it appears that such an approach must be far inferior to the test based on counting process and its mean/rate function model, since the latter utilizes not only the first, but all subsequent event times. Indeed, under the Andersen-Gill multiplicative intensity model assumption (for local alternatives), it is asymptotically semiparametrically efficient, analogously to the log-rank test being efficient for failure time data under the Cox model assumption. This can also be seen intuitively that under the Andersen-Gill multiplicative intensity model, the Fisher information, thus the effective sample size, is proportional to the total number of events that all study subjects experience. The more events are included, the larger Fisher information it results in. With this in mind, it is somewhat surprising to find out that the mean/rate function-based method for recurrent event time data as described in Lin et al. (2) is less efficient than 3

4 the log-rank test with time-to-the-first-event data for testing treatment difference when there is large patient-to-patient variability. The main focus of this paper is to show both analytically and numerically that when the subject-to-subject variability is too high, the mean/rate function based test could be quite inefficient, even less efficient than the log-rank test using a single event time only. Under a quite general setting, asymptotic relative efficiency is expressed analytically, resulting in a threshold-type criterion for efficiency comparison of the two methods. The rest of the paper is organized as follows. In the next section, general notation and setup are described and main theoretical findings are given. Section 3 is devoted to simulation studies, which reinforce the theoretical results. Possible extensions and discussions of the results are given in Section 4. The Appendix contains mathematical derivations of the theoretical results. 2. Main Results In this section, we first introduce the notation and basic assumptions for recurrent event time data. We then describe the two nonparametric tests, one the mean test based on comparing the total number of recurrent events, adjust for censoring, and the other the log-rank test with the first event time. Finally we derive asymptotic efficiencies in general forms as well as under specific model assumptions. 2.. Notation and Assumptions Following Lin et al. (2), let Ni (t) be the number of events experienced by the ith subject during time interval [, t, and Z i an indicator, taking values of or, indicating which of the two treatments is received by the subject. Thus dni (t) = Ni (t) Ni (t ) indicates whether or not an event occurs at t. The censoring time is denoted by C i. In other words, N i (t) = Ni (t C i ) is the observed counting process, where a b = min(a, b). So the observations consist of {N i ( ), C i, Z i ; i =,, n}. Throughout, censoring time C i is assumed to be conditionally 4

5 independent of N i ( ) given Z i. Clearly, the conditional independence allows heterogeneous censorship between the two treatment groups. Unlike the counting processes arising from single failure time data that only take values or, N i can take integer values greater than. Let T ik = inf{t : N i (t) = k}, which defines the occurrence time of the kth event on the ith subject, k and i =,, n. For each i, T ik, k, are censored by a common censoring time C i. So, expressing in terms of event times rather than counting processes, observations consist of {T ik, δ ik, Z i ; k, i =, n}, where T ik = T ik C i, δ ik = I(T ik C i) and I( ) is the usual indicator function. The null hypothesis of no treatment difference between the two comparison groups is tantamount to that the counting processes N i, i =,, n are independent and identically distributed (iid). By definition, it also entails that for each fixed k, Tik, i =,, n are also iid. In particular, the first event times, T i, i =,, n are iid random variables Nonparametric Test Statistics The proportional mean function model specifies that E[N i (t) Z i = e β Z i µ (t), where µ is the baseline mean function and β the true value of the regression parameter. Under this model assumption, an unbiased Cox-type estimating function is U R (β) = i= {Z i Z(β, t)}dn i (t), (2.) where Z(β, t) = n i= I(C i t)e βz i Z i / n i= I(C i t)e βz i. The case of β = corresponds to the null hypothesis of no treatment difference. From the results of Pepe and Cai (993), Lawless and Nadeau(995) and Lin et al. (2), it follows that under the null hypothesis n /2 U R () is asymptotically normal with mean and variance consistently estimated by V = n [ {Z i Z(t)}{dN i (t) I(C i t)dˆµ(t)} i= 5 2

6 where Z(t) = n i= Z ii(c i t)/ n i= I(C i t) and ˆµ(t) = t n i= dn i(s)/ n i= I(C i s). Let U R = U R (). So a two-sided α-level test is to reject the null hypothesis if U 2 R /n V > χ 2 α(), where χ 2 α() is the upper α quantile of the χ 2 () distribution. Again note that validity of this test does not require the proportional mean model assumption. Alternatively, the classical log-rank test using the first event times is also a valid nonparametric test. The log-rank statistic has form U L = i= {Z i Z L (t)}dn L i (t), (2.2) where Z L (t) = n i= Z ii(ti C i t)/ n i= I(T i C i t) and Ni L (t) = δ i I(Ti C i t). Its variance under the null is approximately n i= {Z i Z L (t)} 2 dni L (t), so UL 2/ n i= {Z i Z L (t)} 2 dn L i (t) follows the χ 2 () distribution Efficiency Comparison To compare efficiencies of the preceding two nonparametric tests, it is necessary to specify a parametric or a semiparametric model for the alternatives. To this end, we consider the following frailty model, which can be found in Andersen, Borgan, Gill and Keiding (993) and Lin et al. (2). Let ξ i be iid positive random variables that represent the random effect or frailty. For identifiability, assume Eξ i =. Let σ 2 = V ar(ξ i ). For each i, conditional on Z i and ξ i, counting process Ni (t) is assumed to have compensator ξ i e βz i µ (t). Thus dn i (t) has intensity I(C i t)ξ i e βz i dµ (t). This is a special case of the proportional mean function model since the expectation conditional on C i and Z i only produces I(C i t)e βz i dµ (t). As usual, ξ i incorporates the within-subject dependency. A summary of such dependency is σ 2. For the case of σ 2 =, it reduces to the Anderson-Gill multiplicative intensity model and N i has independent increasements under a monotone transformation of time. A large value for σ 2 indicates a high within-subject correlation. 6

7 Asymptotic relative efficiency (ARE) is evaluated at contiguous alternatives (Hajek and Sidak, 967; Serfling, 98). Let β = β n = b/ n be the parameter value under the frailty model as described. For two test statistics, S and S 2, the ARE of S relative to S 2 is defined to be ARE(S, S 2 ) = e e 2, [E βn (S j ) 2 where e j = lim, j =, 2. n V ar (S j ) where E βn is the expectation taken under the contiguous alternatives, and E and V ar are the expectation and variance taken under the null hypothesis (β = ), respectively. For simplicity, it assumes a balanced allocation, i.e. subjects are evenly divided into the two treatment groups. Let G (t) and G (t) denote the censoring survival distributions for the treatment groups Z i = and Z i =, respectively. Similarly, let S (t) and S (t) denote the survival functions of the first event times of the two treatment groups. And define [ 2 [ Σ R = σ 2 E {Z i µ Z (t)}i(c i t)dµ (t) + E {Z i µ Z (t)} 2 I(C i t)dµ (t), [ A R = E {Z i µ Z (t)} 2 I(C i t)dµ (t), [ Σ L = E {Z i µ L Z(t)} 2 dni L (t), ( [ A L = E {Z i µ L Z(t)} 2 I(T i C i t)d µ (t) E ) ξ{ξe ξµ (t) }, (2.3) E ξ {e ξµ (t) } where E ξ is the expectation with respect to the frailty ξ and µ Z (t) = G (t) G (t) + G (t), µl Z(t) = G (t)s (t) G (t)s (t) + G (t)s (t). Note that in (2.3), µ L Z (t) = µ Z(t) since S (t) = S (t) under the null hypothesis. The following theorem gives the asymptotic relative efficiency of the mean test, U R, using the counting processes data versus the log-rank test, U L, using the first event times only. Theorem The ARE of U R relative to U L can be expressed as ARE(U R, U L ) = A2 R Σ L A 2 L Σ. (2.4) R 7

8 In addition, under the contiguous alternatives β = β n = b/ n, ARE(U R, U L ) can be consistently estimated by (Â2 R ˆΣ L )/(Â2 L ˆΣ R ), where ˆΣ R = σ2 n [ 2 {Z i Z(t)}I(C i t)dˆµ(t) + n i=  L = n [ {Z i Z(t)} 2 I(C i t)dˆµ(t),  R = [ {Z i n Z(t)} 2 I(C i t)dˆµ(t), i= ˆΣ L = [ {Z i n Z L (t)} 2 dni L (t), i= ( [ n {Z i Z L (t)} 2 j= I(T i C i t)d I{N ) j(t) = } n j= I{N. j(t) = } i= i= The proof of Theorem is given in the appendix. Based on Theorem, the log-rank test is more powerful than the mean test in terms of asymptotic relative efficiency if (A 2 R Σ L)/(A 2 L Σ R) <. In practice, we may use (Â2 R ˆΣ L )/(Â2 L ˆΣ R ) <, or equivalently, σ 2 > nâ2 ˆΣ R L /Â2 L n [ i= {Z i Z(t)} 2 I(C i t)dˆµ(t) n [ i= {Z i Z(t)}I(C i t)dˆµ(t) 2 as a threshold dictating that the log-rank test is more efficient. To simplify the ARE(U R, U L ) given in Theorem, we further assume that G (t) = G (t), i.e. the two treatment groups have the common censoring distribution. Thus, µ Z (t) /2 and we have the following Theorem. Theorem 2 Suppose G (t) = G (t) = G(t), then the ARE(U R, U L ) can be expressed as ( E {I(T i C i t)}d E (δ i )[E {µ (C i )} 2 [ µ (t) E ξ{ξe ξµ (t) } E ξ {e ξµ (t) } ) 2 [σ2 E {µ 2 (C i )} + E {µ (C i )}. (2.5) 8

9 Recall that ξ is the frailty with Eξ = and V ar(ξ) = σ 2. When σ 2 =, i.e. ξ or there is no random effect, (2.5) reduces to ARE(U R, U L ) = E {µ (C i )} E {µ (T i C i )} (2.6) since E (δ i ) = E {ξµ (T i C i )} = E {µ (T i C i )}. It is clear that ARE(U R, U L ) >, which is not surprising since U R (β) in this case is a semiparametrically efficient estimating function for β. We now argue that in the other extreme case where σ 2 is large, it is possible that the logrank test (U L ) is more efficient than the mean test (U R ). To avoid technical complication, we will require that the frailty ξ be bounded away from as σ 2 becomes large. Our result is summarized by the following corollary. Corollary Under the same setting as in Theorem 2, suppose in addition that ξ c, where c is a positive constant less than. Then for sufficiently large σ 2, (2.5) approaches. In particular, ARE(U R, U L ) < for all large σ 2. Corollary is more difficult to see from the form of (2.5). A proof is given in the appendix. The constraint imposed on the random effect ξ is to ensure a certain proportion of subjects experience at least one event. This is certainly true if it is a mixture of a constant with a positive random variable. For example, we can take ξ = p + ( p)η, where η is a Gamma random variable with mean and variance σ 2 /( p) 2 with a constant p (, ). Example As an illustration, consider the case in which the censoring time is uniformly distributed over interval [, 3 and the baseline mean function µ (t) = t. The frailty ξ = p+( p)η, 9

10 as being just described. Then it can be shown that the asymptotic relative efficiency ARE(U R, U L ) = 3[ 3 { } 3 +σ 2 σ 2 /( p) 2 e pt dt t/( p) (4σ 2 + 2)[ 3 ( t 3 ){ +σ 2 t/( p) } σ 2 /( p) 2 e pt {p +. p }dt (+σ 2 t/( p)) 2 2 Figure plots the ARE against σ 2 with p =.2. The crossing of the ARE from above to below the horizontal line of agrees with the claim of Corollary. 3. Simulation Results Simulation studies are conducted to compare the efficiencies (power) of the mean test (U R ) and the log-rank test (U L ) to see how the results of Section 2 hold under realistic sample sizes. To this end, we consider the standard two-arm, parallel and balanced design with m = n/2 subjects assigned randomly to each of the two treatment groups, i.e. Z = or with the same probability. Recurrent event times for each subject are generated from the frailty model, which is described in Section 2.3, with µ (t) = t and ξ = p + ( p)η, where p =.2 or.5 and η is a gamma random variable with mean and variance σ 2 η. Different values are chosen for σ 2 η which lead to various choices for the variance σ 2 ξ of ξ. The null hypothesis is β = and alternatives are taken to be β =.5 when m = 2 and β =.8 when m =. Two kinds of followup or censoring time distributions are considered here. In the first case, C is generated from uniform[,3 as in Lin et al. (2), which gives on average.5 events per subject in the control group (Z = ) and.5, 2.47 and 3.34 events per subject for β =,.5 and.8 respectively in the treatment group (Z = ). In the second case, exponential distribution with hazard rate being.8 is considered, which gives on average.25 events per subject in the control group and.25, 2.6 and 2.78 events per subject for β =,.5 and.8 respectively in the treatment group. Given the censoring time C i and the frailty ξ i, the total number of events K i on the ith subject is generated from a Poisson distribution with the mean ξ i e β Z i µ (C i ). In addition, the actual event times (T i, T i2,, T ik i ) are the order statistics of a set of K i independently

11 and identically distributed random variables, which are generated from the following density function (Ross 983) where µ ( ) is the derivative of µ ( ). K i µ (t ij ) f i (t i, t i2,, t i,ki K i, ξ i, C i ) = K i! µ (C i ), j= Results of the simulation studies are summarized in Tables and 2. Each entry in the table was based on simulated data sets. The first table corresponds to the case of uniform censoring time while the second that of exponential censoring time. They strongly support the theoretical findings of Section 2. In particular, both tables clearly show that the mean recurrence test has higher power than the log-rank test does when the variance of ξ is relatively small, which corresponds to small within-subject correlation. On the other hand, the log-rank test becomes more powerful if the variance of ξ is large. The type one errors for both tests are close to their nominal level. Next, as suggested by the referee, we conduct another set of simulations using the positive stable distribution for the frailty. Note that under the positive stable frailty model, time to the first event follows the proportional hazards model (Hougaard, 2), and thus the log-rank test should be most efficient if only the first occurrence times are used. Moreover, the positive stable distribution has variance of infinity. Thus, the mean test tends to loose power. As in the first simulation study, we consider the standard two-arm, parallel and balanced design with m = 5 or, and β =,.5,.8 or.. Recurrence event times are generated from the positive stable frailty model using the similar method as in the first simulation study. Here the positive stable frailties are generated using a R package rstable. The censoring times are generated from uniform[,3. Besides the mean test and the log-rank test for the first occurrence times, we also include the log-rank test based on the second occurrence times for comparison. The simulation

12 results are summarized in Table 3. Based on the simulation results, the type one errors (under β = ) for all three tests are close to their nominal level. But the log-rank test based on the first occurrence times has bigger power than the other two tests under all the alternatives and sample sizes under inquiry, which agree with our expectation. 4. Discussion This paper deals primarily with the issue of efficiency for two competing nonparametric twosample tests with recurrent event time data. It is found that the mean test by using Coxtype partial likelihood score (Andersen and Gill, 982) for the counting process with a robust variance standardization may not be as efficient as one would have expected when there is large variability among study subjects in terms of number of events. Indeed, by formulating the variability using a multiplicative frailty, an analytic expression for the relative efficiency can be derived. In addition, based on the analytical result, a threshold is also constructed to dictate which method to be more efficient. Finally, because of the mean recurrence and time to first event are two different kinds of endpoint, care must be taken in choosing between them so that no misleading interpretation results. In particular, it is only when longer time to first event likely to imply less frequent recurrence that the substitution of the mean test by the log-rank test becomes meaningful. Likewise, if delaying the first event time is the main objective, then the mean test can be used when less frequent recurrence implies longer time to first event. Of course, when the two tests are exchangeable in terms of interpretation, the efficiency will be a main consideration in deciding which one to choose. In addition, it is also interesting to study the optimal combination of the various log-rank tests based on the first, the second and other sequential event times. This will be investigated in our future research. 2

13 Acknowledgement The author would like to thank the Editor Professor Xuming He and the referee for their insightful and constructive comments. The author also thanks Professor Zhiliang Ying for the helpful discussion of the paper. Wenbin Lu s research was partially supported by National Science Foundation Grant DMS Appendix Proof of Theorem : From Serfling(98,.2), we get ARE(U R, U L ) = Γ L Γ R, where Γ K = lim n V ar (n /2 U K )/{E βn (n /2 U K )} 2, K = R, L. Let µ (t Z i ; β n ) be the cumulative hazard function for the first event time T i of the ith subject under the contiguous alternative. Then µ (t Z i ; β n ) = log{p βn (T i > t Z i )} = log[e ξ {P βn (T i > t ξ i )} = log[e ξ {e ξ ie βnz iµ (t) }. For the mean test, lim E β n (n /2 U R ) n [ = lim n E {Z i Z(t)}I(C i t)e β nz i dµ (t) n = lim n n = lim n n i= i= [ E {Z i Z(t)}I(C i t){ + β n Z i + o(β n )}dµ (t) [ E {Z i Z(t)}Z i I(C i t)dµ (t)β n i= [ = be {Z i µ Z (t)} 2 I(C i t)dµ (t) = ba R, 3

14 lim V ar (n /2 U R ) = lim V ar n n = lim n n = lim n n i= {V ar ( E [ + E (V ar [ [ n i= {Z i µ Z (t)}dn i (t) ) {Z i µ Z (t)}dn i (t) ξ i, Z i, C i )} {Z i µ Z (t)}dn i (t) ξ i, Z i, C i {V ar [ξ i {Z i µ Z (t)}i(c i t)dµ (t) i= + E (V ar [ )} {Z i µ Z (t)}{dn i (t) ξ i I(C i t)dµ (t)} ξ i, Z i, C i [ 2 [ = σ 2 E {Z i µ Z (t)}i(c i t)dµ (t) + E {Z i µ Z (t)} 2 I(C i t)dµ (t) = Σ R Therefor Γ R = Σ R /(ba R ) 2. For the log-rank test, we first apply Taylor expansion to the cumulative hazard function µ (t β n ; Z i ) to get µ (t β n ; Z i ) = µ, (t) + Z i β n µ,(t) + o(β n ), where µ, (t) = log[e ξ {e ξµ (t) } and µ,(t) = µ (t) E ξ{ξe ξµ (t) } E ξ {e ξµ (t) }. Thus lim E β n (n /2 U L ) n [ = lim n E {Z i Z L (t)}i(t i C i t)dµ (t Z i ; β n ) n = lim n n = lim n n i= i= [ E {Z i Z L (t)}i(t i C i t){dµ, (t) + Z i β n dµ,(t) + o(β n )} i= ( [ E {Z i Z L (t)} 2 I(T i C i t)β n d µ (t) E ξ{ξe ξµ (t) } ( = be {Z i µ L Z(t)} 2 I(T i C i t)d [ lim V ar (n /2 U L ) = lim V ar n n n [ = E {Z i µ L Z(t)} 2 dni L (t) = Σ L. [ µ (t) E ξ{ξe ξµ (t) } E ξ {e ξµ (t) } i= E ξ {e ξµ (t) } ) = ba L {Z i µ L Z(t)}dN L i (t) ) Hence Γ L = Σ L /(ba L ) 2. Then (2.4) established in Theorem easily follows. In addition, by 4

15 the law of large numbers we have, under the contiguous alternatives, n I{N i (t) = k} {µ (t)} k E ξ {ξ k e ξµ (t) }, k =,. i= as n. Now it is easy to show, also by the law of large numbers, that Σ K and A K (K = R, L) can be consistently estimated by ˆΣ K and ÂK, respectively. Therefore, the remain part of Theorem also holds. Proof of Corollary : First we have that [E {µ (C)} 2 /[σ 2 E {µ 2 (C)} + E {µ (C)} is bounded above by /σ 2. In addition, E (δ i ) = P (T i C i ). Applying integration by part, we have = = [ E {I(T i C i t)}d µ (t) E ξ{ξe ξµ (t) } E ξ {e ξµ (t) } [ G(t)E ξ {e ξµ (t) }d µ (t) E ξ{ξe ξµ (t) } E ξ {e ξµ (t) } µ (t) E ξ{ξe ξµ (t) } E ξ {e ξµ (t) } d[g(t)e ξ{e ξµ (t) } which is equivalent to E ξ {ξe ξµ (t) }µ (t)d{ G(t)} + [E ξ {ξe ξµ (t) } 2 E ξ {e ξµ (t) } µ (t)g(t)dµ (t). Since ξ c, we have E ξ {ξe ξµ (t) } c E ξ {e ξµ (t) }. By Jensen s Inequality, E ξ {e ξµ (t) } e µ (t) since e x is a convex function. Furthermore, we have E ξ {e ξµ (t) } e c µ (t). Thus [ E {I(T i C i t)}d µ (t) E ξ{ξe ξµ (t) } is bounded below by E ξ {e ξµ (t) } Hence c µ (t)e µ (t) d{ G(t)} + c 2 e (2 c )µ (t) µ (t)g(t)dµ (t). ARE(U R, U L ) σ 2 [ c µ (t)e µ (t) d{ G(t)} + c 2 e (2 c )µ (t) µ (t)g(t)dµ (t), 2 which goes to when σ 2 goes to. In particular, for all sufficiently large σ 2, ARE(U R, U L ) is less than. 5

16 References Andersen, P. K., Borgan, O., Gill, R. D. and Keiding, N. (992). Statistical Models Based on Counting Processes. Springer-Verlag. Andersen, P. K. and Gill, R. D. (982). Cox s regression model for counting processes: a large sample study. Ann. Statist., -2. Chang, S-H. and Wang, M-C. (999). Conditional regression analysis for recurrence time data. J. Amer. Statist. Assoc. 94, Cook, R. J., Lawless, J. F. and Nadeau, C. (996). Robust test for treatment comparisons based on recurrent event responses. Biometrics 52, Cox, D. R. (972). Regression models and life tables (with Discussion). J. R. Statist. Soc. B 34, Cox, D. R. (975). Partial likelihood. Biometrika 62, Fleming, T. R. and Harrington, D. P. (99). Counting Processes and Survival Analysis. New York: John Wiley and Sons. Gail, M. H., Santner, T. J. and Brown, C. C. (98). An analysis of comparative carcinogenisis experiments based on multiple times to tumor. Biometrics 36, Gehan, E. A. (965). A generalized Wilcoxon test for comparing arbitrarily singly censored samples. Biometrika 52, Hajek, J. and Sidak, Z. (967). Theory of Rank Tests. Academic Press, New York. Harrington, D. P. and Fleming, T. R. (982). A class of rank test procedures for censored survival data. Biometrika 69,

17 Hougaard, P. (2). Analysis of Multivariate Survival Data. Springer, New York. Kalbfleisch, J. D. and Prentice, R. L. (98). The Statistical Analysis of Failure Time Data. New York: John Wiley and Sons. Lawless, J. F. and Nadeau, C. (995). Some simple robust methods for the analysis of recurrent events. Technometrics 37, Lawless, J. F., Nadeau, C. and Cook, R. J. (997). Analysis of mean and rate functions for recurrent events. In Proceedings of the First Seattle Symposium in Biostatistics: Survival Analysis, Eds. D. Y. Lin and T. R. Fleming, pp New York: Springer-Verlag. Lin, D. Y., Wei, L. J., Yang, I. and Ying, Z. (2). Robust inferences for the Andersen-Gill counting process model. J. R. Statist. Soc. B 62, Mantel, N. and Haenszel, W. (959), Statistical aspects of the analysis of data from retrospective studies of disease, J. Nat. Cancer Inst. 22, Pepe, M. S. & Cai, J. (993). Some graphical displays and marginal regression analyses for recurrent failure times and time dependent covariates. J. Amer. Statist. Assoc. 88, Peto, R. and Peto, J. (972). Asymptotically efficient rank invariant test procedures (with Discussion). J. R. Statist. Soc. A 35, Prentice, R. L. (978). Linear rank tests with right censored data. Biometrika 65, Ross, S. M. (983). Stochastic processes. New York: Wiley. Serfling, R. J. (98). Approximation Theorems of Mathematical Statistics. New York: John Wiley and Sons. 7

18 Wang, M-C. and Chang, S-H. (999). Nonparametric estimation of a recurrent survival function. J. Amer. Statist. Assoc. 94, Wei, L. J., Lin, D. Y. and Weissfeld, L. (989). Regression analysis of multivariate incomplete failure time data by modelling marginal distributions. J. Amer. Statist. Assoc. 84,

19 Table : Simulation results for gamma frailty model under uniform censoring m β p σ 2 η σ 2 ξ POW L (type I error) POW R (type I error) (.58).976 (.52) (.5).927 (.57) (.47).853 (.53) (.49).789 (.6) (.5). (.44) (.45).994 (.48) (.37).974 (.48) (.53).968 (.58) (.56).936 (.53) (.45).848 (.6) (.39).76 (.44) (.43).69 (.5) (.54).997 (.46) (.39).967 (.5) (.56).93 (.55) (.43).99 (.62) σ 2 η and σ 2 ξ are the variances of η and ξ, respectively. POW L and POW R are the powers of the log-rank test and mean test, respectively. Note that ξ = p + ( p)η. 9

20 Table 2: Simulation results for gamma frailty model under exponential censoring m β p ση 2 σξ 2 POW L (type I error) POW R (type I error) (.52).955 (.47) (.49).86 (.57) (.54).794 (.59) (.43).752 (.44) (.55).999 (.54) (.54).98 (.52) (.5).969 (.47) (.52).943 (.49) (.55).874 (.52) (.59).74 (.58) (.42).687 (.6) (.53).66 (.53) (.47).987 (.59) (.48).938 (.43) (.52).9 (.5) (.43).869 (.48) The notations are the same as in Table. 2

21 Table 3: Simulation results for positive stable frailty model under uniform censoring m β POW R POW L POW L POW L 2 is the power of the log-rank test based on the second occurrence times. 2

22 ARE value Variance of the mixed Gamma frailty Figure : Asymptotic relative efficiency curve of the mean test vs. the log-rank test 22

FULL LIKELIHOOD INFERENCES IN THE COX MODEL

FULL LIKELIHOOD INFERENCES IN THE COX MODEL October 20, 2007 FULL LIKELIHOOD INFERENCES IN THE COX MODEL BY JIAN-JIAN REN 1 AND MAI ZHOU 2 University of Central Florida and University of Kentucky Abstract We use the empirical likelihood approach

More information

Tests of independence for censored bivariate failure time data

Tests of independence for censored bivariate failure time data Tests of independence for censored bivariate failure time data Abstract Bivariate failure time data is widely used in survival analysis, for example, in twins study. This article presents a class of χ

More information

Other Survival Models. (1) Non-PH models. We briefly discussed the non-proportional hazards (non-ph) model

Other Survival Models. (1) Non-PH models. We briefly discussed the non-proportional hazards (non-ph) model Other Survival Models (1) Non-PH models We briefly discussed the non-proportional hazards (non-ph) model λ(t Z) = λ 0 (t) exp{β(t) Z}, where β(t) can be estimated by: piecewise constants (recall how);

More information

On the Breslow estimator

On the Breslow estimator Lifetime Data Anal (27) 13:471 48 DOI 1.17/s1985-7-948-y On the Breslow estimator D. Y. Lin Received: 5 April 27 / Accepted: 16 July 27 / Published online: 2 September 27 Springer Science+Business Media,

More information

Hypothesis Testing Based on the Maximum of Two Statistics from Weighted and Unweighted Estimating Equations

Hypothesis Testing Based on the Maximum of Two Statistics from Weighted and Unweighted Estimating Equations Hypothesis Testing Based on the Maximum of Two Statistics from Weighted and Unweighted Estimating Equations Takeshi Emura and Hisayuki Tsukuma Abstract For testing the regression parameter in multivariate

More information

GOODNESS-OF-FIT TESTS FOR ARCHIMEDEAN COPULA MODELS

GOODNESS-OF-FIT TESTS FOR ARCHIMEDEAN COPULA MODELS Statistica Sinica 20 (2010), 441-453 GOODNESS-OF-FIT TESTS FOR ARCHIMEDEAN COPULA MODELS Antai Wang Georgetown University Medical Center Abstract: In this paper, we propose two tests for parametric models

More information

Part III. Hypothesis Testing. III.1. Log-rank Test for Right-censored Failure Time Data

Part III. Hypothesis Testing. III.1. Log-rank Test for Right-censored Failure Time Data 1 Part III. Hypothesis Testing III.1. Log-rank Test for Right-censored Failure Time Data Consider a survival study consisting of n independent subjects from p different populations with survival functions

More information

A COMPARISON OF POISSON AND BINOMIAL EMPIRICAL LIKELIHOOD Mai Zhou and Hui Fang University of Kentucky

A COMPARISON OF POISSON AND BINOMIAL EMPIRICAL LIKELIHOOD Mai Zhou and Hui Fang University of Kentucky A COMPARISON OF POISSON AND BINOMIAL EMPIRICAL LIKELIHOOD Mai Zhou and Hui Fang University of Kentucky Empirical likelihood with right censored data were studied by Thomas and Grunkmier (1975), Li (1995),

More information

Statistics 262: Intermediate Biostatistics Non-parametric Survival Analysis

Statistics 262: Intermediate Biostatistics Non-parametric Survival Analysis Statistics 262: Intermediate Biostatistics Non-parametric Survival Analysis Jonathan Taylor & Kristin Cobb Statistics 262: Intermediate Biostatistics p.1/?? Overview of today s class Kaplan-Meier Curve

More information

University of California, Berkeley

University of California, Berkeley University of California, Berkeley U.C. Berkeley Division of Biostatistics Working Paper Series Year 24 Paper 153 A Note on Empirical Likelihood Inference of Residual Life Regression Ying Qing Chen Yichuan

More information

Chapter 7 Fall Chapter 7 Hypothesis testing Hypotheses of interest: (A) 1-sample

Chapter 7 Fall Chapter 7 Hypothesis testing Hypotheses of interest: (A) 1-sample Bios 323: Applied Survival Analysis Qingxia (Cindy) Chen Chapter 7 Fall 2012 Chapter 7 Hypothesis testing Hypotheses of interest: (A) 1-sample H 0 : S(t) = S 0 (t), where S 0 ( ) is known survival function,

More information

Power and Sample Size Calculations with the Additive Hazards Model

Power and Sample Size Calculations with the Additive Hazards Model Journal of Data Science 10(2012), 143-155 Power and Sample Size Calculations with the Additive Hazards Model Ling Chen, Chengjie Xiong, J. Philip Miller and Feng Gao Washington University School of Medicine

More information

Lecture 5 Models and methods for recurrent event data

Lecture 5 Models and methods for recurrent event data Lecture 5 Models and methods for recurrent event data Recurrent and multiple events are commonly encountered in longitudinal studies. In this chapter we consider ordered recurrent and multiple events.

More information

Lecture 12. Multivariate Survival Data Statistics Survival Analysis. Presented March 8, 2016

Lecture 12. Multivariate Survival Data Statistics Survival Analysis. Presented March 8, 2016 Statistics 255 - Survival Analysis Presented March 8, 2016 Dan Gillen Department of Statistics University of California, Irvine 12.1 Examples Clustered or correlated survival times Disease onset in family

More information

STAT 331. Accelerated Failure Time Models. Previously, we have focused on multiplicative intensity models, where

STAT 331. Accelerated Failure Time Models. Previously, we have focused on multiplicative intensity models, where STAT 331 Accelerated Failure Time Models Previously, we have focused on multiplicative intensity models, where h t z) = h 0 t) g z). These can also be expressed as H t z) = H 0 t) g z) or S t z) = e Ht

More information

log T = β T Z + ɛ Zi Z(u; β) } dn i (ue βzi ) = 0,

log T = β T Z + ɛ Zi Z(u; β) } dn i (ue βzi ) = 0, Accelerated failure time model: log T = β T Z + ɛ β estimation: solve where S n ( β) = n i=1 { Zi Z(u; β) } dn i (ue βzi ) = 0, Z(u; β) = j Z j Y j (ue βz j) j Y j (ue βz j) How do we show the asymptotics

More information

Survival Analysis for Case-Cohort Studies

Survival Analysis for Case-Cohort Studies Survival Analysis for ase-ohort Studies Petr Klášterecký Dept. of Probability and Mathematical Statistics, Faculty of Mathematics and Physics, harles University, Prague, zech Republic e-mail: petr.klasterecky@matfyz.cz

More information

Published online: 10 Apr 2012.

Published online: 10 Apr 2012. This article was downloaded by: Columbia University] On: 23 March 215, At: 12:7 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 172954 Registered office: Mortimer

More information

Rank Regression Analysis of Multivariate Failure Time Data Based on Marginal Linear Models

Rank Regression Analysis of Multivariate Failure Time Data Based on Marginal Linear Models doi: 10.1111/j.1467-9469.2005.00487.x Published by Blacwell Publishing Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA Vol 33: 1 23, 2006 Ran Regression Analysis

More information

FULL LIKELIHOOD INFERENCES IN THE COX MODEL: AN EMPIRICAL LIKELIHOOD APPROACH

FULL LIKELIHOOD INFERENCES IN THE COX MODEL: AN EMPIRICAL LIKELIHOOD APPROACH FULL LIKELIHOOD INFERENCES IN THE COX MODEL: AN EMPIRICAL LIKELIHOOD APPROACH Jian-Jian Ren 1 and Mai Zhou 2 University of Central Florida and University of Kentucky Abstract: For the regression parameter

More information

TESTS FOR LOCATION WITH K SAMPLES UNDER THE KOZIOL-GREEN MODEL OF RANDOM CENSORSHIP Key Words: Ke Wu Department of Mathematics University of Mississip

TESTS FOR LOCATION WITH K SAMPLES UNDER THE KOZIOL-GREEN MODEL OF RANDOM CENSORSHIP Key Words: Ke Wu Department of Mathematics University of Mississip TESTS FOR LOCATION WITH K SAMPLES UNDER THE KOIOL-GREEN MODEL OF RANDOM CENSORSHIP Key Words: Ke Wu Department of Mathematics University of Mississippi University, MS38677 K-sample location test, Koziol-Green

More information

4. Comparison of Two (K) Samples

4. Comparison of Two (K) Samples 4. Comparison of Two (K) Samples K=2 Problem: compare the survival distributions between two groups. E: comparing treatments on patients with a particular disease. Z: Treatment indicator, i.e. Z = 1 for

More information

STAT Sample Problem: General Asymptotic Results

STAT Sample Problem: General Asymptotic Results STAT331 1-Sample Problem: General Asymptotic Results In this unit we will consider the 1-sample problem and prove the consistency and asymptotic normality of the Nelson-Aalen estimator of the cumulative

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

Frailty Models and Copulas: Similarities and Differences

Frailty Models and Copulas: Similarities and Differences Frailty Models and Copulas: Similarities and Differences KLARA GOETHALS, PAUL JANSSEN & LUC DUCHATEAU Department of Physiology and Biometrics, Ghent University, Belgium; Center for Statistics, Hasselt

More information

Quantile Regression for Residual Life and Empirical Likelihood

Quantile Regression for Residual Life and Empirical Likelihood Quantile Regression for Residual Life and Empirical Likelihood Mai Zhou email: mai@ms.uky.edu Department of Statistics, University of Kentucky, Lexington, KY 40506-0027, USA Jong-Hyeon Jeong email: jeong@nsabp.pitt.edu

More information

A comparison study of the nonparametric tests based on the empirical distributions

A comparison study of the nonparametric tests based on the empirical distributions 통계연구 (2015), 제 20 권제 3 호, 1-12 A comparison study of the nonparametric tests based on the empirical distributions Hyo-Il Park 1) Abstract In this study, we propose a nonparametric test based on the empirical

More information

A Poisson Process Approach for Recurrent Event Data with Environmental Covariates NRCSE. T e c h n i c a l R e p o r t S e r i e s. NRCSE-TRS No.

A Poisson Process Approach for Recurrent Event Data with Environmental Covariates NRCSE. T e c h n i c a l R e p o r t S e r i e s. NRCSE-TRS No. A Poisson Process Approach for Recurrent Event Data with Environmental Covariates Anup Dewanji Suresh H. Moolgavkar NRCSE T e c h n i c a l R e p o r t S e r i e s NRCSE-TRS No. 028 July 28, 1999 A POISSON

More information

Semiparametric Regression

Semiparametric Regression Semiparametric Regression Patrick Breheny October 22 Patrick Breheny Survival Data Analysis (BIOS 7210) 1/23 Introduction Over the past few weeks, we ve introduced a variety of regression models under

More information

Improving Efficiency of Inferences in Randomized Clinical Trials Using Auxiliary Covariates

Improving Efficiency of Inferences in Randomized Clinical Trials Using Auxiliary Covariates Improving Efficiency of Inferences in Randomized Clinical Trials Using Auxiliary Covariates Anastasios (Butch) Tsiatis Department of Statistics North Carolina State University http://www.stat.ncsu.edu/

More information

STAT331. Cox s Proportional Hazards Model

STAT331. Cox s Proportional Hazards Model STAT331 Cox s Proportional Hazards Model In this unit we introduce Cox s proportional hazards (Cox s PH) model, give a heuristic development of the partial likelihood function, and discuss adaptations

More information

STAT 331. Martingale Central Limit Theorem and Related Results

STAT 331. Martingale Central Limit Theorem and Related Results STAT 331 Martingale Central Limit Theorem and Related Results In this unit we discuss a version of the martingale central limit theorem, which states that under certain conditions, a sum of orthogonal

More information

Models for Multivariate Panel Count Data

Models for Multivariate Panel Count Data Semiparametric Models for Multivariate Panel Count Data KyungMann Kim University of Wisconsin-Madison kmkim@biostat.wisc.edu 2 April 2015 Outline 1 Introduction 2 3 4 Panel Count Data Motivation Previous

More information

Pairwise rank based likelihood for estimating the relationship between two homogeneous populations and their mixture proportion

Pairwise rank based likelihood for estimating the relationship between two homogeneous populations and their mixture proportion Pairwise rank based likelihood for estimating the relationship between two homogeneous populations and their mixture proportion Glenn Heller and Jing Qin Department of Epidemiology and Biostatistics Memorial

More information

Efficiency of Profile/Partial Likelihood in the Cox Model

Efficiency of Profile/Partial Likelihood in the Cox Model Efficiency of Profile/Partial Likelihood in the Cox Model Yuichi Hirose School of Mathematics, Statistics and Operations Research, Victoria University of Wellington, New Zealand Summary. This paper shows

More information

Full likelihood inferences in the Cox model: an empirical likelihood approach

Full likelihood inferences in the Cox model: an empirical likelihood approach Ann Inst Stat Math 2011) 63:1005 1018 DOI 10.1007/s10463-010-0272-y Full likelihood inferences in the Cox model: an empirical likelihood approach Jian-Jian Ren Mai Zhou Received: 22 September 2008 / Revised:

More information

CHAPTER 1 A MAINTENANCE MODEL FOR COMPONENTS EXPOSED TO SEVERAL FAILURE MECHANISMS AND IMPERFECT REPAIR

CHAPTER 1 A MAINTENANCE MODEL FOR COMPONENTS EXPOSED TO SEVERAL FAILURE MECHANISMS AND IMPERFECT REPAIR CHAPTER 1 A MAINTENANCE MODEL FOR COMPONENTS EXPOSED TO SEVERAL FAILURE MECHANISMS AND IMPERFECT REPAIR Helge Langseth and Bo Henry Lindqvist Department of Mathematical Sciences Norwegian University of

More information

On robust and efficient estimation of the center of. Symmetry.

On robust and efficient estimation of the center of. Symmetry. On robust and efficient estimation of the center of symmetry Howard D. Bondell Department of Statistics, North Carolina State University Raleigh, NC 27695-8203, U.S.A (email: bondell@stat.ncsu.edu) Abstract

More information

Likelihood ratio confidence bands in nonparametric regression with censored data

Likelihood ratio confidence bands in nonparametric regression with censored data Likelihood ratio confidence bands in nonparametric regression with censored data Gang Li University of California at Los Angeles Department of Biostatistics Ingrid Van Keilegom Eindhoven University of

More information

PhD course in Advanced survival analysis. One-sample tests. Properties. Idea: (ABGK, sect. V.1.1) Counting process N(t)

PhD course in Advanced survival analysis. One-sample tests. Properties. Idea: (ABGK, sect. V.1.1) Counting process N(t) PhD course in Advanced survival analysis. (ABGK, sect. V.1.1) One-sample tests. Counting process N(t) Non-parametric hypothesis tests. Parametric models. Intensity process λ(t) = α(t)y (t) satisfying Aalen

More information

Analysis of transformation models with censored data

Analysis of transformation models with censored data Biometrika (1995), 82,4, pp. 835-45 Printed in Great Britain Analysis of transformation models with censored data BY S. C. CHENG Department of Biomathematics, M. D. Anderson Cancer Center, University of

More information

Linear rank statistics

Linear rank statistics Linear rank statistics Comparison of two groups. Consider the failure time T ij of j-th subject in the i-th group for i = 1 or ; the first group is often called control, and the second treatment. Let n

More information

STAT331. Combining Martingales, Stochastic Integrals, and Applications to Logrank Test & Cox s Model

STAT331. Combining Martingales, Stochastic Integrals, and Applications to Logrank Test & Cox s Model STAT331 Combining Martingales, Stochastic Integrals, and Applications to Logrank Test & Cox s Model Because of Theorem 2.5.1 in Fleming and Harrington, see Unit 11: For counting process martingales with

More information

A Simulation Study on Confidence Interval Procedures of Some Mean Cumulative Function Estimators

A Simulation Study on Confidence Interval Procedures of Some Mean Cumulative Function Estimators Statistics Preprints Statistics -00 A Simulation Study on Confidence Interval Procedures of Some Mean Cumulative Function Estimators Jianying Zuo Iowa State University, jiyizu@iastate.edu William Q. Meeker

More information

Longitudinal + Reliability = Joint Modeling

Longitudinal + Reliability = Joint Modeling Longitudinal + Reliability = Joint Modeling Carles Serrat Institute of Statistics and Mathematics Applied to Building CYTED-HAROSA International Workshop November 21-22, 2013 Barcelona Mainly from Rizopoulos,

More information

UNIVERSITY OF CALIFORNIA, SAN DIEGO

UNIVERSITY OF CALIFORNIA, SAN DIEGO UNIVERSITY OF CALIFORNIA, SAN DIEGO Estimation of the primary hazard ratio in the presence of a secondary covariate with non-proportional hazards An undergraduate honors thesis submitted to the Department

More information

EMPIRICAL ENVELOPE MLE AND LR TESTS. Mai Zhou University of Kentucky

EMPIRICAL ENVELOPE MLE AND LR TESTS. Mai Zhou University of Kentucky EMPIRICAL ENVELOPE MLE AND LR TESTS Mai Zhou University of Kentucky Summary We study in this paper some nonparametric inference problems where the nonparametric maximum likelihood estimator (NPMLE) are

More information

Simulation-based robust IV inference for lifetime data

Simulation-based robust IV inference for lifetime data Simulation-based robust IV inference for lifetime data Anand Acharya 1 Lynda Khalaf 1 Marcel Voia 1 Myra Yazbeck 2 David Wensley 3 1 Department of Economics Carleton University 2 Department of Economics

More information

Chapter 2 Inference on Mean Residual Life-Overview

Chapter 2 Inference on Mean Residual Life-Overview Chapter 2 Inference on Mean Residual Life-Overview Statistical inference based on the remaining lifetimes would be intuitively more appealing than the popular hazard function defined as the risk of immediate

More information

ST745: Survival Analysis: Nonparametric methods

ST745: Survival Analysis: Nonparametric methods ST745: Survival Analysis: Nonparametric methods Eric B. Laber Department of Statistics, North Carolina State University February 5, 2015 The KM estimator is used ubiquitously in medical studies to estimate

More information

MARGINAL REGRESSION MODELS FOR RECURRENT AND TERMINAL EVENTS

MARGINAL REGRESSION MODELS FOR RECURRENT AND TERMINAL EVENTS Statistica Sinica 12(22), 663-688 MARGINAL REGRESSION MODELS FOR RECURRENT AND TERMINAL EVENTS Debashis Ghosh and D. Y. Lin University of Michigan and University of North Carolina Abstract: A major complication

More information

asymptotic normality of nonparametric M-estimators with applications to hypothesis testing for panel count data

asymptotic normality of nonparametric M-estimators with applications to hypothesis testing for panel count data asymptotic normality of nonparametric M-estimators with applications to hypothesis testing for panel count data Xingqiu Zhao and Ying Zhang The Hong Kong Polytechnic University and Indiana University Abstract:

More information

Survival Analysis. Lu Tian and Richard Olshen Stanford University

Survival Analysis. Lu Tian and Richard Olshen Stanford University 1 Survival Analysis Lu Tian and Richard Olshen Stanford University 2 Survival Time/ Failure Time/Event Time We will introduce various statistical methods for analyzing survival outcomes What is the survival

More information

Panel Count Data Regression with Informative Observation Times

Panel Count Data Regression with Informative Observation Times UW Biostatistics Working Paper Series 3-16-2010 Panel Count Data Regression with Informative Observation Times Petra Buzkova University of Washington, buzkova@u.washington.edu Suggested Citation Buzkova,

More information

11 Survival Analysis and Empirical Likelihood

11 Survival Analysis and Empirical Likelihood 11 Survival Analysis and Empirical Likelihood The first paper of empirical likelihood is actually about confidence intervals with the Kaplan-Meier estimator (Thomas and Grunkmeier 1979), i.e. deals with

More information

ANALYSIS OF COMPETING RISKS DATA WITH MISSING CAUSE OF FAILURE UNDER ADDITIVE HAZARDS MODEL

ANALYSIS OF COMPETING RISKS DATA WITH MISSING CAUSE OF FAILURE UNDER ADDITIVE HAZARDS MODEL Statistica Sinica 18(28, 219-234 ANALYSIS OF COMPETING RISKS DATA WITH MISSING CAUSE OF FAILURE UNDER ADDITIVE HAZARDS MODEL Wenbin Lu and Yu Liang North Carolina State University and SAS Institute Inc.

More information

GROUPED SURVIVAL DATA. Florida State University and Medical College of Wisconsin

GROUPED SURVIVAL DATA. Florida State University and Medical College of Wisconsin FITTING COX'S PROPORTIONAL HAZARDS MODEL USING GROUPED SURVIVAL DATA Ian W. McKeague and Mei-Jie Zhang Florida State University and Medical College of Wisconsin Cox's proportional hazard model is often

More information

Department of Mathematical Sciences, Norwegian University of Science and Technology, Trondheim

Department of Mathematical Sciences, Norwegian University of Science and Technology, Trondheim Tests for trend in more than one repairable system. Jan Terje Kvaly Department of Mathematical Sciences, Norwegian University of Science and Technology, Trondheim ABSTRACT: If failure time data from several

More information

Non-parametric Tests for the Comparison of Point Processes Based on Incomplete Data

Non-parametric Tests for the Comparison of Point Processes Based on Incomplete Data Published by Blackwell Publishers Ltd, 108 Cowley Road, Oxford OX4 1JF, UK and 350 Main Street, Malden, MA 02148, USA Vol 28: 725±732, 2001 Non-parametric Tests for the Comparison of Point Processes Based

More information

Goodness-of-Fit Tests With Right-Censored Data by Edsel A. Pe~na Department of Statistics University of South Carolina Colloquium Talk August 31, 2 Research supported by an NIH Grant 1 1. Practical Problem

More information

Concepts and Tests for Trend in Recurrent Event Processes

Concepts and Tests for Trend in Recurrent Event Processes JIRSS (2013) Vol. 12, No. 1, pp 35-69 Concepts and Tests for Trend in Recurrent Event Processes R. J. Cook, J. F. Lawless Department of Statistics and Actuarial Science, University of Waterloo, Ontario,

More information

PENALIZED LIKELIHOOD PARAMETER ESTIMATION FOR ADDITIVE HAZARD MODELS WITH INTERVAL CENSORED DATA

PENALIZED LIKELIHOOD PARAMETER ESTIMATION FOR ADDITIVE HAZARD MODELS WITH INTERVAL CENSORED DATA PENALIZED LIKELIHOOD PARAMETER ESTIMATION FOR ADDITIVE HAZARD MODELS WITH INTERVAL CENSORED DATA Kasun Rathnayake ; A/Prof Jun Ma Department of Statistics Faculty of Science and Engineering Macquarie University

More information

A Multivariate Two-Sample Mean Test for Small Sample Size and Missing Data

A Multivariate Two-Sample Mean Test for Small Sample Size and Missing Data A Multivariate Two-Sample Mean Test for Small Sample Size and Missing Data Yujun Wu, Marc G. Genton, 1 and Leonard A. Stefanski 2 Department of Biostatistics, School of Public Health, University of Medicine

More information

Analysis of recurrent gap time data using the weighted risk-set. method and the modified within-cluster resampling method

Analysis of recurrent gap time data using the weighted risk-set. method and the modified within-cluster resampling method STATISTICS IN MEDICINE Statist. Med. 29; :1 27 [Version: 22/9/18 v1.11] Analysis of recurrent gap time data using the weighted risk-set method and the modified within-cluster resampling method Xianghua

More information

A Regression Model For Recurrent Events With Distribution Free Correlation Structure

A Regression Model For Recurrent Events With Distribution Free Correlation Structure A Regression Model For Recurrent Events With Distribution Free Correlation Structure J. Pénichoux(1), A. Latouche(2), T. Moreau(1) (1) INSERM U780 (2) Université de Versailles, EA2506 ISCB - 2009 - Prague

More information

Sample size re-estimation in clinical trials. Dealing with those unknowns. Chris Jennison. University of Kyoto, January 2018

Sample size re-estimation in clinical trials. Dealing with those unknowns. Chris Jennison. University of Kyoto, January 2018 Sample Size Re-estimation in Clinical Trials: Dealing with those unknowns Christopher Jennison Department of Mathematical Sciences, University of Bath, UK http://people.bath.ac.uk/mascj University of Kyoto,

More information

Applications of Basu's TheorelTI. Dennis D. Boos and Jacqueline M. Hughes-Oliver I Department of Statistics, North Car-;'lina State University

Applications of Basu's TheorelTI. Dennis D. Boos and Jacqueline M. Hughes-Oliver I Department of Statistics, North Car-;'lina State University i Applications of Basu's TheorelTI by '. Dennis D. Boos and Jacqueline M. Hughes-Oliver I Department of Statistics, North Car-;'lina State University January 1997 Institute of Statistics ii-limeo Series

More information

Monotonicity and Aging Properties of Random Sums

Monotonicity and Aging Properties of Random Sums Monotonicity and Aging Properties of Random Sums Jun Cai and Gordon E. Willmot Department of Statistics and Actuarial Science University of Waterloo Waterloo, Ontario Canada N2L 3G1 E-mail: jcai@uwaterloo.ca,

More information

Survival Analysis I (CHL5209H)

Survival Analysis I (CHL5209H) Survival Analysis Dalla Lana School of Public Health University of Toronto olli.saarela@utoronto.ca January 7, 2015 31-1 Literature Clayton D & Hills M (1993): Statistical Models in Epidemiology. Not really

More information

Modelling and Analysis of Recurrent Event Data

Modelling and Analysis of Recurrent Event Data Modelling and Analysis of Recurrent Event Data Edsel A. Peña Department of Statistics University of South Carolina Research support from NIH, NSF, and USC/MUSC Collaborative Grants Joint work with Prof.

More information

Lecture 3. Truncation, length-bias and prevalence sampling

Lecture 3. Truncation, length-bias and prevalence sampling Lecture 3. Truncation, length-bias and prevalence sampling 3.1 Prevalent sampling Statistical techniques for truncated data have been integrated into survival analysis in last two decades. Truncation in

More information

Cox s proportional hazards model and Cox s partial likelihood

Cox s proportional hazards model and Cox s partial likelihood Cox s proportional hazards model and Cox s partial likelihood Rasmus Waagepetersen October 12, 2018 1 / 27 Non-parametric vs. parametric Suppose we want to estimate unknown function, e.g. survival function.

More information

The Proportional Hazard Model and the Modelling of Recurrent Failure Data: Analysis of a Disconnector Population in Sweden. Sweden

The Proportional Hazard Model and the Modelling of Recurrent Failure Data: Analysis of a Disconnector Population in Sweden. Sweden PS1 Life Cycle Asset Management The Proportional Hazard Model and the Modelling of Recurrent Failure Data: Analysis of a Disconnector Population in Sweden J. H. Jürgensen 1, A.L. Brodersson 2, P. Hilber

More information

Survival Analysis Math 434 Fall 2011

Survival Analysis Math 434 Fall 2011 Survival Analysis Math 434 Fall 2011 Part IV: Chap. 8,9.2,9.3,11: Semiparametric Proportional Hazards Regression Jimin Ding Math Dept. www.math.wustl.edu/ jmding/math434/fall09/index.html Basic Model Setup

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

1 Introduction. 2 Residuals in PH model

1 Introduction. 2 Residuals in PH model Supplementary Material for Diagnostic Plotting Methods for Proportional Hazards Models With Time-dependent Covariates or Time-varying Regression Coefficients BY QIQING YU, JUNYI DONG Department of Mathematical

More information

Typical Survival Data Arising From a Clinical Trial. Censoring. The Survivor Function. Mathematical Definitions Introduction

Typical Survival Data Arising From a Clinical Trial. Censoring. The Survivor Function. Mathematical Definitions Introduction Outline CHL 5225H Advanced Statistical Methods for Clinical Trials: Survival Analysis Prof. Kevin E. Thorpe Defining Survival Data Mathematical Definitions Non-parametric Estimates of Survival Comparing

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

Size and Shape of Confidence Regions from Extended Empirical Likelihood Tests

Size and Shape of Confidence Regions from Extended Empirical Likelihood Tests Biometrika (2014),,, pp. 1 13 C 2014 Biometrika Trust Printed in Great Britain Size and Shape of Confidence Regions from Extended Empirical Likelihood Tests BY M. ZHOU Department of Statistics, University

More information

Estimation of the Bivariate and Marginal Distributions with Censored Data

Estimation of the Bivariate and Marginal Distributions with Censored Data Estimation of the Bivariate and Marginal Distributions with Censored Data Michael Akritas and Ingrid Van Keilegom Penn State University and Eindhoven University of Technology May 22, 2 Abstract Two new

More information

Survival Distributions, Hazard Functions, Cumulative Hazards

Survival Distributions, Hazard Functions, Cumulative Hazards BIO 244: Unit 1 Survival Distributions, Hazard Functions, Cumulative Hazards 1.1 Definitions: The goals of this unit are to introduce notation, discuss ways of probabilistically describing the distribution

More information

Estimation and Inference of Quantile Regression. for Survival Data under Biased Sampling

Estimation and Inference of Quantile Regression. for Survival Data under Biased Sampling Estimation and Inference of Quantile Regression for Survival Data under Biased Sampling Supplementary Materials: Proofs of the Main Results S1 Verification of the weight function v i (t) for the lengthbiased

More information

Statistical Inference of Covariate-Adjusted Randomized Experiments

Statistical Inference of Covariate-Adjusted Randomized Experiments 1 Statistical Inference of Covariate-Adjusted Randomized Experiments Feifang Hu Department of Statistics George Washington University Joint research with Wei Ma, Yichen Qin and Yang Li Email: feifang@gwu.edu

More information

Link to published article: (Access to content may be restricted)

Link to published article:   (Access to content may be restricted) Kvaløy, J.T. (2002) Covariate Order Tests for Covariate Effect. Lifetime Data Analysis, 8(1), pp. 35-51 Link to published article: http://link.springer.com/article/10.1023/a:1013518815447 (Access to content

More information

Attributable Risk Function in the Proportional Hazards Model

Attributable Risk Function in the Proportional Hazards Model UW Biostatistics Working Paper Series 5-31-2005 Attributable Risk Function in the Proportional Hazards Model Ying Qing Chen Fred Hutchinson Cancer Research Center, yqchen@u.washington.edu Chengcheng Hu

More information

A Measure of Association for Bivariate Frailty Distributions

A Measure of Association for Bivariate Frailty Distributions journal of multivariate analysis 56, 6074 (996) article no. 0004 A Measure of Association for Bivariate Frailty Distributions Amita K. Manatunga Emory University and David Oakes University of Rochester

More information

Integrated likelihoods in survival models for highlystratified

Integrated likelihoods in survival models for highlystratified Working Paper Series, N. 1, January 2014 Integrated likelihoods in survival models for highlystratified censored data Giuliana Cortese Department of Statistical Sciences University of Padua Italy Nicola

More information

Plugin Confidence Intervals in Discrete Distributions

Plugin Confidence Intervals in Discrete Distributions Plugin Confidence Intervals in Discrete Distributions T. Tony Cai Department of Statistics The Wharton School University of Pennsylvania Philadelphia, PA 19104 Abstract The standard Wald interval is widely

More information

Regularization in Cox Frailty Models

Regularization in Cox Frailty Models Regularization in Cox Frailty Models Andreas Groll 1, Trevor Hastie 2, Gerhard Tutz 3 1 Ludwig-Maximilians-Universität Munich, Department of Mathematics, Theresienstraße 39, 80333 Munich, Germany 2 University

More information

Statistical Analysis of Competing Risks With Missing Causes of Failure

Statistical Analysis of Competing Risks With Missing Causes of Failure Proceedings 59th ISI World Statistics Congress, 25-3 August 213, Hong Kong (Session STS9) p.1223 Statistical Analysis of Competing Risks With Missing Causes of Failure Isha Dewan 1,3 and Uttara V. Naik-Nimbalkar

More information

On least-squares regression with censored data

On least-squares regression with censored data Biometrika (2006), 93, 1, pp. 147 161 2006 Biometrika Trust Printed in Great Britain On least-squares regression with censored data BY ZHEZHEN JIN Department of Biostatistics, Columbia University, New

More information

Competing risks data analysis under the accelerated failure time model with missing cause of failure

Competing risks data analysis under the accelerated failure time model with missing cause of failure Ann Inst Stat Math 2016 68:855 876 DOI 10.1007/s10463-015-0516-y Competing risks data analysis under the accelerated failure time model with missing cause of failure Ming Zheng Renxin Lin Wen Yu Received:

More information

A Recursive Formula for the Kaplan-Meier Estimator with Mean Constraints

A Recursive Formula for the Kaplan-Meier Estimator with Mean Constraints Noname manuscript No. (will be inserted by the editor) A Recursive Formula for the Kaplan-Meier Estimator with Mean Constraints Mai Zhou Yifan Yang Received: date / Accepted: date Abstract In this note

More information

Statistics and Probability Letters. Using randomization tests to preserve type I error with response adaptive and covariate adaptive randomization

Statistics and Probability Letters. Using randomization tests to preserve type I error with response adaptive and covariate adaptive randomization Statistics and Probability Letters ( ) Contents lists available at ScienceDirect Statistics and Probability Letters journal homepage: wwwelseviercom/locate/stapro Using randomization tests to preserve

More information

Issues on quantile autoregression

Issues on quantile autoregression Issues on quantile autoregression Jianqing Fan and Yingying Fan We congratulate Koenker and Xiao on their interesting and important contribution to the quantile autoregression (QAR). The paper provides

More information

Accelerated Failure Time Models: A Review

Accelerated Failure Time Models: A Review International Journal of Performability Engineering, Vol. 10, No. 01, 2014, pp.23-29. RAMS Consultants Printed in India Accelerated Failure Time Models: A Review JEAN-FRANÇOIS DUPUY * IRMAR/INSA of Rennes,

More information

Analysis of Gamma and Weibull Lifetime Data under a General Censoring Scheme and in the presence of Covariates

Analysis of Gamma and Weibull Lifetime Data under a General Censoring Scheme and in the presence of Covariates Communications in Statistics - Theory and Methods ISSN: 0361-0926 (Print) 1532-415X (Online) Journal homepage: http://www.tandfonline.com/loi/lsta20 Analysis of Gamma and Weibull Lifetime Data under a

More information

Comparing Distribution Functions via Empirical Likelihood

Comparing Distribution Functions via Empirical Likelihood Georgia State University ScholarWorks @ Georgia State University Mathematics and Statistics Faculty Publications Department of Mathematics and Statistics 25 Comparing Distribution Functions via Empirical

More information

The Design of a Survival Study

The Design of a Survival Study The Design of a Survival Study The design of survival studies are usually based on the logrank test, and sometimes assumes the exponential distribution. As in standard designs, the power depends on The

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