Skip to content

[R-meta] N of trials or participants?

4 messages · Patrizio E Tressoldi, Wolfgang Viechtbauer

Original
Patrizio E Tressoldi
#
Il 31/01/2023 21:51, Viechtbauer, Wolfgang (NP)wrote:
Yes, I mean a standardized mean difference or a difference between 
proportions. In many studies, participants' results are based on a 
different number of trials.

Patrizio


Subject: [R-meta] N of trials or participants?

When and how is it possible to use N of trials instead of N of

participants in the standardized effect sizes estimation?

It is likely that this question already had a response. In this case,

please send me links or references.

Thanks
Wolfgang Viechtbauer
# 
I think you are using 'trials' in a different way than how I interpreted it. I understood it as meaning 'studies' (as in 'clinical trials') but you seem to be referring to repeated measurements on individual subjects. Can you describe in more detail what kind of data you are thinking of?

Best,
Wolfgang
Patrizio E Tressoldi
#
Il 01/02/2023 08:44, Viechtbauer, Wolfgang (NP) wrote:
Sorry, as trials I mean number of repetitions of the task, e.g., to 
guess or identify a target as in many perceptual or cognitive 
psychological tasks.

Patrizio
Wolfgang Viechtbauer
# 
Thanks for clarifying.

So let's say each subject goes through m trials on some task and either correctly identifies a target or not on a given trial. Let x_j denote the number of trials on which the target was correctly identified by the jth subject and so p_j = x_j/m would be the corresponding proportion of trials. It would be unusual that p_j is reported for the N subjects in the study. Typically, one might have the mean proportion (so: sum p_j / N) and the standard deviation of the proportions. If the subjects are actually in two groups/conditions, then one can just compute a standardized mean difference in the usual manner based on the two means and SDs. The number of trials (m) is not relevant for this.

Best,
Wolfgang