Sas proc phreg

For each cohort of adult females, mean blood feeding frequency (=mean of total number of meals each female took divided by the number of meals each female was offered), mean number of blood meals, mean lifetime fecundity for those females laying eggs, mean length of the first egg production cycle, and mean fecundity in the first cycle for those females laying eggs, were analyzed using randomized complete block ANOVA with replicate and treatment as variables (PROC GLM, SAS Institute Inc. 2002–2005 ). Adult longevity (=days from eclosion until death) was analyzed using Cox proportional hazards survival analysis (SAS PROC PHREG, SAS Institute Inc. 2002–2005 ) including block, and treatment as class variables and female wing length as a covariate. We also included cumulative reproductive output in the model as time-dependent covariates (i.e., their values can change over the course of the experiment, See Leisnham et al., 2008 (link)). This time dependent covariate tests for a physiological cost of reproduction in our laboratory environment. Estimated age-specific hazard rate was compared among treatments using contrasts in this proportional hazard model. Hazard functions (mean hazard rate at age x) were calculated for each treatment in PHREG by using the BASELINE statement.

In two previous studies (i.e., Forand & DeRubeis, 2014 (link); Brouwer et al., 2019 (link)), the authors examined a series of models
to investigate dysfunctional attitudes and variables characterizing
different types of extreme responding as predictors of relapse. In this
study, we planned to examine those same models. We conducted a total of four
separate Cox proportional hazard models to test the relationship between the
variables of interest and risk of relapse. The predictors of the specific
models are as follows: (1) the DAS total score, (2) the DAS controlling for
style responses, (3) the DAS controlling for content responses, and (4) the
ratio of endorsing more positive extreme responses on style versus content
items while controlling for the number of positive extreme responses.
Patients who were lost to follow-up or changed treatments were censored at
the time of the event. Prior to running our main models, we also examined
the following covariates: prior depressive episodes, residual symptoms
(measured by PHQ-9 at posttreatment), and study segment (i.e., phase 1
immediate, phase 1 delayed, and phase 2 immediate as reported in Schmidt et al., 2018 (link)). Analyses were
performed with SAS PROC PHREG. For ease of interpretation,
we standardized (M = 0, SD = 1) all
predictors prior to running the main analyses.

Time-varying covariates including exposures are common in survival analyses. A Cox model can be modified to allow for time-varying covariates. For example, a time-varying indicator variable for wait time to heart transplantation was included in a modified partial likelihood to estimate the effect of heart transplant on hazard of death with wait time treated as unexposed and post-transplantation time treated as exposed time. ^{25 ,26} When the frequency of change in the covariate value is not small, a more flexible approach called Cox model for counting process ^{27 -29} can be used. This model allows for recurring events. In the counting process model, multiple observations for each individual are created. Each observation reflects a unique time interval and includes 1) the number of events an individual experiences during the time interval; 2) an indicator variable for an individual being at risk during the time interval; and 3) corresponding values of covariates including the exposure variable during the time interval. Survival data in counting processes formulation can be modeled with existing statistical software (e.g., SAS PROC PHREG). ³⁰