Hi all,

I have unbalanced panel data from 2010-2017 of people who participated in a screening program. Every two years, the program actively invites people to come and screen until they reach the age of 69. However, after the age of 69, people could continue using the screening program but through self-referral.

I am interested in identifying the factors associated with using the screening program through self-referral. The data include observations from individuals
  1. who used the screening program before and after the age of 69
  2. who used the screening program before the age of 69 only and didn’t continue using the screening program after they reached the age of 69. E.g. no self-referral
  3. who used the screening program only before the age of 69 because they haven’t yet reached the age limit of 69 during the study period
I have generated a binary outcome variable to indicate if people used the screening program before and after age of 69 or if they only used the screening program before the age of 69. E.g. 1= screened before and after the age of 69 and 0= screened only before the age of 69

The number of observation per individual range from 1 to 8.

I am currently using a random effect logistic regression an example of the code is below:

xt id
xtlogit outcome i.independent1 i. independent2 i. independent3 i. independent4 I, re or

Appreciate your advice if xtlogit is the right model to use? Also, when I run xtlogit, I get an output but there is ‘not concave' message next to some of the iterations and I am not sure why.

Thank you!