Logistic Regression on Large Set of Binary Data

[KC]

Hello,

I'm quite new to R, so please forgive me in advance if this problem seems simple or if I'm leaving something important out.

We have run two experiments in which participants were shown a subset of 160 ambiguous scenarios and asked to give an interpretation of the number of entities involved in each scenario (singular vs. plural). What we are trying to do is find out if the responses in experiment N (32 participants) can be used to predict the responses in experiment T (45 participants). All responses have been converted to a binary scale, where 1 represents a response of plural and 0 represents a response of singular. Participants did not all see the same stimuli as participants saw one of 2 lists in experiment N and they saw one of 3 lists in experiment T. However, none of the stimuli included in the dataset were used in one experiment but not the other (i.e. there are responses from both experiments for each item).

Currently, the data are arranged as shown:

Item#	Experiment	Participant	Response
1	N	1	1
2	N	1	1
3	N	1	na
.	.	.	.
.	.	.	.
160	N	1	na
1	N	2	1
2	N	2	na
3	N	2	na
.	.	.	.
.	.	.	.
160	N	2	1
1	T	1	1
2	T	1	1
3	T	1	na
.	.	.	.
.	.	.	.
160	T	1	0
1	T	2	1
2	T	2	1
3	T	2	1
.	.	.	.
.	.	.	.
160	T	2	1

Again, we want to know if responses for experiment N predict those for experiment T. Additionally, this dataset is quite large (~4900 responses, 1 response per row) and it will not display in its entirety on the R workspace.

If anyone has any input regarding this matter, it would be greatly appreciated.

Thank you!
KC

[Stefan Th. Gries]

Why not run a binary logistic regression on N (with glm(..., family=binomial)) and apply it to T (with predict(..., newdata=...))?
STG
--
Stefan Th. Gries
-----------------------------------------------
University of California, Santa Barbara
http://www.linguistics.ucsb.edu/faculty/stgries
-----------------------------------------------