4
$\begingroup$

My data consists in two samples relative to the expression of a given gene, with 4 replications in both samples.

I am interested in testing whether the two samples come from the same population, taking into account the fact that I have repeated measures by design.

I've been reading about different non-parametric approaches and the Friedman's test seems to be one option. However, given the extremely low sample size, I am not quite sure that this test can be adopted.

Can you suggest if I can proceed this way? Can/should other approaches be adopted?

EDIT

alternatively, would any exact test would be ok for my specific need?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
1
$\begingroup$

In my case I analysed the results about the abundance of the genes (qPCR) and when the my dataset was not normal ditribution, I applied the Kruskal-Wallis test and if I saw a significant different apply the pairwise.wilcox.test. I hope to help you.

Improve this answer
$\endgroup$
5
  • $\begingroup$ thanks @Giorgia, which sample size would be ok for it (taking into account that I have 4+4 repeated meauses per gene)? $\endgroup$
    – Stefano Lombardi
    Jun 30, 2015 at 9:16
  • $\begingroup$ In my case I have 4 replicate for each treatment, for each time point. In total I have 10 treatments and 6 time points. Do you have differente time to sampling or not? $\endgroup$
    – Giorgia
    Jun 30, 2015 at 12:47
  • $\begingroup$ I need a little more information about your dataset. $\endgroup$
    – Giorgia
    Jun 30, 2015 at 12:53
  • $\begingroup$ thanks, in my case I have two datasets relative to the same set of genes, and 4 replicates for each treatment. So, basically 2 treatments x 4 time points. $\endgroup$
    – Stefano Lombardi
    Jun 30, 2015 at 20:51
  • $\begingroup$ You create a table with the treatment, time and abundance of your gene. In this table you put the control and the treatment. You create this same table for each treatment. After you save the table in .csv and you import by R. You calculate the Shapiro test and in function of this you apply the ANOVA or the Kruskal-Wallis test. If the first time you have a non-normality distribution you try to transform the data. $\endgroup$
    – Giorgia
    Jul 1, 2015 at 6:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.