Help needed with data analysis

Paul

New Member
#1
Hi,

I am a optometrist in practice and I have done a small study of the effect of a treatment on other clinical measures that aren't the main target of the treatment. Basically, I have take pre and post measurements and want to know how best to analyse my data to determine if there is a significant effect on the measures from the treatment.

My sample size is small (n=11) as this was only an indicative study based on clinical observations. I have 19 different measures for each patient that I want to assess change.

I have tried t-tests but I don't think it is a correction assumption that the data in the general population is normally distributed for this subset of patients (they have a specific condition). The paired t-tests basically show no statistical significance across most measures, although this contradicts the clinical observations in general and even in my sample as 10/11 patients measurements improved in what I would call clinically significant ways, including resolution of symptoms.

I have looked at non-parametric tests such as Wilcoxan? to compare the before and after tests but I am unsure how to use it. I haven't had to use stats for 10 years since I was at university, so I've forgotten much of what I did know.

Any help appreciated.

Paul
 

Paul

New Member
#3
Further to this, it appears from my research and the above post that the Wilcoxon Matched Pairs Sign Rank Test is appropriate. But is it when I want to compare the before and after effects of the treatment on about 15 variables? Does doing multiple WMPSR Tests increase the chance of Type 1? errors (i.e. my p values need to be a lot less to show change in medians?) or is it appropriate to do a WMPSR test on each variable? If so, what is the appropriate test then?

The variables are all measures of visual function but don't necessarily relate to each other.

Also, by clinically looking at the totality of the variables, an optometrist can make a clinical judgment if the patient is normal or abnormal. Can I assign 0 and 1 to this and do WMPSR test on the pre and post treatment results to see if the treatment made more subjects normal?

And can I also make a judgment if the clinical profile improved overall (=1) or remained the same (=0) and do WMPSR test on this?

Thanks!