1

Say I conduct a study where i measure physiological signal of a set of N subjects, to record a particular event. I am interested of the duration and the amplitude of that event. The study ends when I recorded M events per subject.

I need M events per subject to analyse the intra-subject variability. I need N subjects to analyse the subject population variability.

So, how can i know the minimum number of N subject and M event I need to record to get a statistically significant study ?

from here i understand that it depends on several factors, but I would like to be able to compute those number for that particular exemple. The distributions will likely be Normal or a bit skewed.

EDIT: may be this is not clear. I need to compare events in one individuals and events among individuals. But here, i have a sample of individuals that are them selfs composed of a sample of events. How i am supposed to sum up individuals to compare them ? do i even need to sum them up (i could just put all events i the same bag) ? And thus i wonder how N and M may influance each orther ?

Hattori
  • 41
  • 2
  • 1
    You've added the [tag:statistical-power] tag. I suggest you take a look at some of the top questions with that tag to learn more about power analysis (which is what you seem to be after). – mkt Jun 01 '21 at 17:03
  • yes, i am also looking at these notion. I understand how the power of a test can vary depending on the difference in means of underlying distribution and the sample variance and size. But what disturb me is that, in my example, the event of interest can be different among people, or among one individual. In other work, i have a sample of subject that, them selfs are composed of sample of events. How i am supposed to sum up individuals to compare them ? do i even need to sum them up (i could just put all events i the same bag) ? I guess this depends of what i am looking for but i am confused. – Hattori Jun 03 '21 at 12:58
  • This is a good question that gets at important issues with sample size calculation, power analysis, and inference in [tag:mixed-model]s. I recommend looking at that tag and this great FAQ: https://bbolker.github.io/mixedmodels-misc/glmmFAQ.html – mkt Jun 03 '21 at 20:29
  • Here's a closely related question: https://stats.stackexchange.com/questions/48374/sample-size-calculation-for-mixed-models and this part of that FAQ is probably most relevant to you: https://bbolker.github.io/mixedmodels-misc/glmmFAQ.html#power-analysis – mkt Jun 03 '21 at 20:32
  • Though you might want to read a bit about the basics of mixed models first, in case you are unfamiliar with them. – mkt Jun 03 '21 at 20:32
  • thanks a lot. i'll dive into it ! – Hattori Jun 04 '21 at 11:13

0 Answers0