Response Feature Analysis--Word List Recall

jcr

New Member
#1
Hi,
I think I'm looking at response feature analysis, but I'm not sure how to go about this. FYI, this post refers only to providing descriptive results--no inferences are to be made.

I am involved in a pilot study involving 2 groups of subjects (n=5 each) who will be given a 10-item list of words to memorize followed by free recall, over a series of 10 trials (same word list). I am interested in somehow describing each subject's improvement over time using a single descriptor. The easy solution, to me, is to just calculate a slope of #correct against #trials for each subject. However, there are a few issues that may prevent this from being reasonable.

1) Some subjects may peak (achieve 100% recall) at trial 5 (for example), meaning trials 6-10 could theoretically remain at that peak of 100%. So this is not linear, or at least potentially linear with a plateau at the tail.

2) Some subjects may peak at trial 10, which would likely be linear, so not really a problem by itself.

3) Some subjects may not reach 100% at trial 10, and we won't know if they ever would.

Any suggestions on how to handle this?

Also, if I am able to calculate a slope or other descriptor for each subject, would it be reasonable to calculate a mean of slopes (or other measure) to describe the overall rate of improvement for each group?

Thank you in advance for any help!
 

Karabiner

TS Contributor
#2
It is a bit surprising that you have to invent a method by yourself,
since the experimental design looks like it has been in use for a
century or more. Isn't there anything about this problem found
in the literature?

Anyway, since aquisition isn't always improving linearly with time/trial,
perhaps #correct = b0 + b1*trial+b2*trial² could be used.

Probably they have better (logarithmic?) models in growth curve analysis.

Just my 2pence

K.