Logistic Regression Stata: ommitted due to collinearity ?? Help

#1
Hi there,

I'm an MSc student in Medical Entomology and I'm in the midst of analysing data for my final project write up. My project is assessing the efficacy of different insecticide treatments applied in experimental huts against malaria mosquitoes. Now my dependent variables are: mortality, blood feeding, exiting rates and the number inside the net. I'm using Stata to analyse this data using logistic regression with the blog it command. The command I have used is as follows:

. xi: blogit tot_dead total i.treat i.hut i.sleeper i.week

This command works by comparing the lowest numbered treatment to the remaining treatments. You gradually drop out the lowest numbered treatment to compare the next one to the remainder using a replace command as follows:

replace treat=9 if treat==0
. xi: blogit tot_dead total i.treat i.hut i.sleeper i.week, nolog or

As you can see my main independent variable is the treatment then I have other variables involved such as hut, sleeper, week. When I did a practice run on this data at the 4 week point it seemed to work fine and came out with all the p values I needed. I have tried to do it again at 6 weeks and Stata has omitted all of the treatments from the model due to 'collinearity'. I'm no statistician and don't really understand the ins and outs of all this so I have no idea how to make it work. Strangely, when I tried to analyse the old data set that worked using the same command, it also omitted all the treatments due to 'collinearity'. I apologise in advance if none of this makes any sense, however any help would be greatly appreciated.
 
#2
Collinearity is a data/sample problem; however, if you realize that there seems to be perfect relationship from the regression, I suggest that you try to run auxiliary regressions first and thereafter, re-run the entire equation. This is just a suggestion to test the variables in the regression.