Hey,
as the title says I'm looking for a logistic regression with repeated measuring. what I found was a generalized estimates equation (gee), which seems to be limited to a large sample size. But there are authors how claim, that a small sample is possible too, like: http://onlinelibrary.wiley.com/doi/10.1002/sim.6817/epdf, they developed a packet for r, called: geesmv. Is that the way you would do it too? Or is there another way?
As the title says this is for a forensic sample, we are obtaining risk levels for violence offenders who are already in treatment and we'd like to test if a given risklevel (before treatment) increases the probability for an offender to get into the treatment.
Samplesize per group will bei arround 30-35. We planning of having n = 70 total
Times measured: k = 2
This is for an expose, so we haven't got the data yet.
Thanks,
m
Hi Maximilian,
Maybe permutation tests could help you. There is an R package called logregperm, here the description from the manual:
A permutation test is used for inference in logistic regression. The procedure is useful when parameter estimates in ordinary logistic regression fail to converge or are unreliable due to small sample size, or when the conditioning in exact conditional logistic regression restricts the sample space too severely
Source: http://www2.uaem.mx/r-mirror/web/packages/logregperm/logregperm.pdf
The package is not in CRAN but you can install it from github here:
https://github.com/cran/logregperm
Hey Angel, thanks very much. I will definitely check that out! Have you ever tried it yourself?
Is there any reason not to model this as GLMM with random intercepts and effects for each pariticipant? The lme4 package for R is pretty handy. You can, of course, also use a GEE - the difference is largely whether you want population estimates or subject-level estimates. In any case, using more than one approach would give you a sensitivity test.
You can refer this book :https://www.researchgate.net/publication/329999817_Guideline_In_Environmental_Forensic_For_Enforcement/stats
For logistic regression analysis a bigger sample size is recommended:
Article Sample Size Guidelines for Logistic Regression from Observat...
If you have longitudinal data, with a single outcome variable, then you can use time-to-event analysis with time-dependent covariates.
- Badges
- Science topic
- Similar topics
- Computer Science
- Program