clarification of time-series analysis
Entering edit mode
Assa Yeroslaviz ★ 1.5k
Last seen 7 weeks ago

In a great post previously answered by @Michael Love I have found his response as such

"strainmut.minute15" is the difference between Mut vs WT at minute 15, controlling for baseline. If you add "strainmutvs_wt" to this, you get the LFC for Mutant vs WT at minute 15, not controlling for baseline. So the second one is the observed difference at minute 15 between the two groups (because you added in the change that was present at time=0).

I have seen this answer and similar in different posts.

What I would like to better understand is the highlighted part. What is the exactly difference between

results(dds, contrast="strainmut.minute15")


results(dds, contrast=list(c("strain_mut_vs_wt","strainmut.minute15"))

what does it means controlling for baseline?



deseq2 baseline LRT • 722 views
Entering edit mode
Last seen 1 hour ago
United States

The first is the difference between the strains at minute 15, controlling for the difference present in the strains at the first time point (t0). That is what we mean by "controlling at baseline".

The second is the sum of the baseline difference and the difference between the strains at minute 15. If you think about four groups of samples:

A: wt at t0 B: mut at t0 C: wt at t1 D mut at t1,

Then the above is giving you (B-A) + ((D-C) - (B-A)). Which gives you D-C, i.e. the difference between Mutant and WT at minute 15, not controlling for baseline. Typically, you want to control for differences at t0 though. I should probably clarify that post to make that more clear. The person was asking for a strange type of analysis, and I gave the answer, but that is not recommended or the usual way to analyze time series data.

Entering edit mode

again thanks. It starts to becoming clearer. But than why should someone use the second design? Is there an event where not controlling for baseline differences is needed?

Entering edit mode

Hmm, I think we're going down an unnecessary path. There was a thread where someone asked me about a strange analysis, but I don't want to be recommending that. I don't know why that person wanted to do that particular analysis, not controlling for baseline. I think it's a bad choice.


Login before adding your answer.

Traffic: 326 users visited in the last hour
Help About
Access RSS

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6