Hello all,
I am counting specific deletion events in the genome. I have a table that looks like this.
Sample Event-1-count All-event-counts
x1 500 500,000
x2 6000 460,000
I would like to compare two samples and more than two samples.
What tests would you recommend?
I was thinking of the two-sample independent T-Test and the Anova.
You are analyzing counts. So use a count model (negative-binomial or quasi-Poisson). "All-even-counts" should be included as an offset.
I'd advice against t-tests/ANOVA because the rates are close to zero so the normal approximation (even of log-rates) may not be good. Apart from that: there exist models specifically for the analysis of count data, and you have count data... but I repeat myself :)
you can additionally see the binomial exact test and the multinomial test.
There are several statistical tests that can be used to compare counts and percentages of counts across samples. Here are a few commonly used ones:
It is important to choose the appropriate test based on the type of data and the research question being addressed.
The suggestions here are excellent if your only consideration is the formal testing. But, if you are looking to do some initial informal thinking about what is going on with lots of samples, you could think of doing some visual plots of your data ... and for this you could look at some of the variance-stabilisation theory. Thus, for "counts", you could do plots using the square-root transformation, while for "percentages" you could use an inverse-sine transformation.
- Badges
- Science method