Hello all,
I am using GEE to estimate my dependent variable. The dependent variable has a lower bound of 0 (its observed value, not censored or truncated) and can take on larger values as well. However, in my dataset, there are a lot of zeros for the dependent variable (about 80% of the time). Would it be acceptable to run a linear GEE model here? (assuming that I probe my results using alternative approaches). From what I understand, GEE is a quasi-likelihood estimator and it has weaker distributional assumptions, so my thoughts were that this would be okay, but I'd be interested in hearing others thoughts. Though to be clear, I am interested in using and defending this approach for my analysis purposes.
Thank you in advance!
Related Posts with GEE and distributional assumptions
Problem with stcox hazard modelHi all, I am trying to use stcox to create a multivariable cox proportional hazard model. I am runni…
Trend analysisI am interested in comparing a linear trend model with 1) power trend, 2) exponential trend, and 3) …
Trouble with nested option in synthHi all, I am trying to use the user-contributed synth to estimate with synthetic control method. I a…
What's the principle behind twoway graph's legend and its label optionDear Stata users, I have an interesting finding regards to twoway graph's legend option, put it pre…
Expand value within groupDear all, I am currently working on a patient consultations database from 2016 to 2021. My purpose …
Subscribe to:
Post Comments (Atom)
0 Response to GEE and distributional assumptions
Post a Comment