3 replies to this topic

[Paja]

Hi,
I would like to ask for some advice with how to statistically analyze my data.
I have 3 specimens from 25 patients taken at the beginning of their disease, when they clinically improve and at the time of release from hospital. In all three specimens I measured the expression of gene of interest. Now I want to know whether there is a sex-deppendent difference in the disease course represented by differences in gene of interest expression. In other words, I would like to analyze differences among times of sampling in male and female patients.
The only way how I am able to do that is to count median (non-Normal distribution) for all three sampling times for men and women and compare these 3 and 3 median values by Mann-Whitney statistics. But this is too simplified I guess and I am loosing much of variability in each group. What would be the proper way to statistically compare my data?

Thank you in advance for any suggestion.
Paja

[mike.sh]

Try using ANOVA with 'sex' and 'disease state' as factors. As there is not equal number of men and women you can either discard extra samples and perform two-way ANOVA or check out ANOVA with unbalanced design.

[bob1]

ANOVA is only for normally distributed populations - don't use it for your analysis unless you can prove or assume normality in the distribution.  You probably want the Kruskal-Wallace test (non-parametric equivalent of the ANOVA) with some sort of post-hoc testing.

[NicoBxl]

for post-hoc use a wilcoxson test