I've got a question about how to do statistical analyses:

 

Let's say a randomized controlled study of the efficacy of a new antiretroviral agent to be used in treating HIV/AIDS --specifically, the one here: http://www.scienceforums.net/forum/showthread.php?t=32796 -- is commenced in HIV+ mice, w/ 20 mice randomly assigned to the treatment group, and 20 mice randomly assigned to the placebo group. 18 mice in the treatment group end up w/ viral loads < 50 copies/ml --which is definied in the literature as clinically insignificant--, and a CD4+ T-cell count of > 500 cells/ul, which is that of a healthy individual --the remaning two succumbed from causes unrelated to HIV--. Those in the placebo group, all died.

 

Now, what can I do with this data, or do I need more data? What data do I need to calclate the mean and standard deviation, so I can calculate the 99% CI --or the 1% p-value; any difference? What CL or p-value would be appropriate in this situation?-- and the effect size? Would the above RCT's results be statistically significant --they sure are clinically significant!--?

 

Thanks!!!

Because the trial contains mice that died of other causes, I would probably analyse this as a markov process and determine rates rather than probabilities. The rates can then be turned into probabilities

 

Probability of survival to time t = exp[-integrate between t and 0 mu(s) ds], where mu(t) is the death rate at time t

 

Maximum likelihood estmates of rates and the variances are probably easily calculated from exposure times and no of deaths