monicathieu
commented
5 years ago ➤ Jessica Younger commented:
Whoa realizing it's been a while. I think I wasn't quite understanding the proposal. I think I miscommunicated this one - we want there to be an arrangement comparison (subitizing, random, grouped) in addition to number of groups, but those arrangement and number of groups shouldn't be crossed, because as you say, there will be too few trials in each bin. And then since there's three levels of grouping, we'd need to do a similar thing for cost where we have subitizing vs grouped, subitizing vs random, and grouped vs random. We don't need subitizing 3 groups vs random 3 groups. Does that make sense?
Question: to do this, I'm looking at modifying the behavior of the underlying summary stats calculation functions to be able to accept more than 2 grouping columns (e.g. arrangement, number_groups, AND the usual correct_button to subset RT summary stats by response accuracy). if this were to happen, would you want the output to contain the following?
Subsetting by 1 x 2 x 3 will basically never be useful, as I suspect there'll never be enough trials in such small bins to make the stats useful in any way. Given that, whenever there are 3 subsetting conditions, one of them is probably response accuracy, I can also tell the code to not cross accuracy with the other conditions, or something. let me know!