Question: How to evaluate correlation for multiple measurements per point?
0
gravatar for max.rempel2
4 months ago by
max.rempel20 wrote:

I've got a question about statistics. I have an experiment in which X - is a length of DNA fragment and for each X, there are 4 measurements of Y. Suppose, it is not obvious that there is an increase of Y with X. What is the proper statistic to verify that this increase is not a fluke?

P.S. I have many experiments like that so the answer will be useful. I don't think a simple correlation will do since the measurements are not independent: there are 4 replicates for each X. Also averaging is not good since this way I would lose a lot of stat power.

sample data

statistics • 194 views
ADD COMMENTlink modified 3 months ago • written 4 months ago by max.rempel20

I would think more along the regression route in order to check the rate of change of one group per unit change in another (<- the beta coefficient). You may try a LOESS or linear fit. Looking at your plot, the increase appears linear. The penultimate measurement looks 'all over the place', though. Note that there are no statisticians here (that I know) - for a better answer, try Cross Validated.

ADD REPLYlink written 4 months ago by Kevin Blighe48k

Thank you very much! Great help! I have similar feedback from a statistician, with similar sentiments.

ADD REPLYlink written 4 months ago by max.rempel20

Yo should be aware of position-related sequencing artifacts

ADD REPLYlink written 4 months ago by chen1.9k

Thank you very much!

ADD REPLYlink written 4 months ago by max.rempel20
0
gravatar for max.rempel2
3 months ago by
max.rempel20 wrote:

Finally, I just used a simple correlation test and it seemed to work just fine.

ADD COMMENTlink written 3 months ago by max.rempel20
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 2191 users visited in the last hour