search

kristabh / gaze-following-analysis

Multi-lab infant gaze following project
MIT License
2 stars 0 forks source link

d_first_meta: why remove subgroups with fewer than 10 babies? #34

Closed AlexisBlack2g closed 5 years ago

AlexisBlack2g commented 5 years ago

I don't think this particular restriction is in the pre-reg, right? Is it from the MB1 or MB1B analyses? Doing this does remove one problem in the data - we currently have data from not mono and not bilingual babies as well as 3-6 and 9-12 mo babies. I believe all of these cases, however, have fewer than 10 subs, so are removed from the meta-analysis.

kristabh commented 5 years ago

Yeah, this is from MB1/B. The rationale there was that if there are groups with fewer than 10, then the estimate of the meta-analytic effect size is unstable. I think we could go either way. You're right that we didn't pre-reg anything about this, probably because we didn't expect any labs would end up with fewer than 10 since the target was a min of 16.

kristabh commented 5 years ago

But age and language-based exclusions should be taken care of in the paper/exclusions script. This gets called in the 03_exclusions script, but only when it gets knitted. It's kind of confusing, but it was set up that way in the mb1 script which I based this off of. So make sure to run the paper/exclusions script prior to running the 03_exclusions script to exclude babies properly.

AlexisBlack2g commented 5 years ago

Ah, thanks - I thought I'd seen the exclusions filtering these out! I ran the chunks in 03_exclusions, but maybe that didn't call in the child exclusions script (or I managed to write-out the wrong file, or something- will check on it)