Suppose an investigator planned repeated measures on n individuals with K time points but the sample size changes due to death event after each time point. The investigator seeks to predict the cause of death by using continuous variables measured at each time point at each individual. how to calculate sample size and what will be the appropriate statistical analysis method for this kind of research problem?
I don't have information about the specific variables, their distribution, or the base-rate for death, however, a first stab might be to use a logistic regression or discriminant analysis to predict the dichotomous death event from prior data available on the continuous variables you mention. I would then be interested in three hit-rates (correct classification as "dead," "not dead," and total). A cross-validated estimate (Lachenbruch U, or "Leave-one-out" type) would be most convincing to me.
It seems that the data that are available differ as K increases, and that may affect the accuracy of said models. Moreover, as less able/resilient individuals may die earlier, it may be that the nature of the model changes due to that. My first notion would be to use a data-available strategy at each K, and then examine the accuracy, but also the relative variable contributions as K increases.
My best to your research,
John
Sayed Mohammed Zeeshan . This sounds like you want to do a survival analysis, where you look at the effect of each of your covariates on time to death. the Time Dependent Cox model allows for survival analysis with repeated measures of covariates. See for example https://cran.r-project.org/web/packages/survival/vignettes/timedep.pdf
or https://stats.stackexchange.com/questions/210609/how-to-use-time-dependent-covariates-with-cox-regression-in-r
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis