I aim to say that there is a certain sensitivity (let say>80%) for people to be sick if X>10. (my actual intent is to check the sensitivity for different values of X)

its a retrospective cohort and the estimated incidence of sick people in the whole group is 3.5%

how or which tool should i use to calculate the sample size? i have tried to find an answer online (including this good chinese site: http://www.statstodo.com/StatsToDoIndex.php#Sample Size) but so far no success

thanks so much for your help! ]]>

I am working with an occupation cohort exposed to radiation and have a dataset containing a number of variables on each worker, including their job title, year of birth, radiation exposure level, and the date on which the radiation sample was collected. Some of the radiation exposure measurements are missing, however, as soon in the attached figure.

One of the methods we wish to try is to multiply impute the missing radiation data using available information provided by the other variables.

My concern: for a given worker, there may be multiple radiation measurements collected in a given year and over the years (measurements were collected whenever the worker performed a certain task, which does not follow a set pattern). Thus, I can expect some correlation between samples on the same worker within a year and between years.

I am using the MICE package in R. However, I am having trouble locating information on how to correctly handle the imputation of a variable with repeated measures on the same individual and the expected correlation. I know MICE can be used for longitudinal data but I am not clear on what, if anything, I need to specify in my code in order to perform the imputations correctly.

I do appreciate the help!

Pam