I want to know if there are differences in firing rate of one neuron between two types of trials: left and right. The data are non-normal, paired and unmatched and the size of samples is different.

For example, I want to compare 15 firing rate for left trials with 11 firing rate for right trials. If I had the same size of trials, I could use Wilcoxon, but it's not the case.

Do you know what test should I choose?

  • 2
    $\begingroup$ Could you please elaborate on how data could be "paired" yet have different sample sizes? $\endgroup$ – whuber May 17 '18 at 13:26
  • $\begingroup$ If there are a different number of trials, how can they be paired? ** sorry, just saw whuber already asked this question $\endgroup$ – rw2 May 17 '18 at 13:33

It seems strange that your data is paired, but you have different sample sizes - maybe you measured firing rate for left and right trials for 11 of the samples, but only left trials for the remaining 4 samples?

If that is the case you could think about using a mixed-effects model. Firing rate would be your response, and you'd be testing whether left/right was a significant predictor. If some of your data is paired, you could include sample (i.e. pair) as a random effect.

You mention that your data are non-normal, but don't elaborate. It may be that a generalized linear mixed effects model would be most appropriate. e.g. if your response is count data, you could specify a poisson distribution.

  • $\begingroup$ Thanks for your response. I checked normality of data with Shapiro test, histograms shapes and q-q plots. Although number of trials are different for left and right, every firing rate belong to the same neuron. So, it means they are paired, right? $\endgroup$ – Ana Fernandez May 17 '18 at 15:20
  • $\begingroup$ And what about applying a Wilcoxon test by adding NaNs to right conditions until getting the same size as left condition? So I would have the same size of samples $\endgroup$ – Ana Fernandez May 17 '18 at 16:37
  • $\begingroup$ It really doesn’t sound like paired data to me, but I don’t know anything about firing neurons and am struggling to understand your experimental design. Maybe you could add a comment to explain it in more detail? Is it the same neuron for all measurements, or just pairs of measurement? $\endgroup$ – rw2 May 19 '18 at 14:01

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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