# How to compare multi-row averages in JMP?

July 22, 2007 2:40 PM Subscribe

How do I compare averages in JMP (statistical software)? Help me stop using Excel.

I have a behavioral data file where each row is an individual response time measurement. Each row contains three fields:

I'm pretty certain that what I'm supposed to do here is a repeated measures ANOVA. For each subject-stimulus pair, obtain an RT average, thus ending up with (n subjects * n stims) averages. I then do the ANOVA on that.

So, what I've been doing is using an Excel Pivottable to collapse/average across subject-stimulus, and then taking the resulting data into JMP for the ANOVA. This seems really silly to me - there's got to be a way to do this all in JMP, probably in just one or two steps, without all this copying and pasting, if I just knew what I was doing.

How might I do that?

I have a behavioral data file where each row is an individual response time measurement. Each row contains three fields:

*subject*(numeric, nominal);

*stimulus*(character, nominal);

*RT*(numeric, continuous). The file contains hundreds of rows per subject (equal number of stimuli-rows per subject), and dozens of subjects.

I'm pretty certain that what I'm supposed to do here is a repeated measures ANOVA. For each subject-stimulus pair, obtain an RT average, thus ending up with (n subjects * n stims) averages. I then do the ANOVA on that.

So, what I've been doing is using an Excel Pivottable to collapse/average across subject-stimulus, and then taking the resulting data into JMP for the ANOVA. This seems really silly to me - there's got to be a way to do this all in JMP, probably in just one or two steps, without all this copying and pasting, if I just knew what I was doing.

How might I do that?

Also, googling "jmp anova repeated measures" pulls up relevant things at #1 and #2.

posted by a robot made out of meat at 4:19 PM on July 22, 2007

posted by a robot made out of meat at 4:19 PM on July 22, 2007

Best answer: If I understand you correctly, I think you can do this with a summary table.

Here's the step-by-step:

1. From the raw data table, choose Table > Summary.

2. Click on RT and under "Statistics" select mean.

3. Drag "Subject" to the "Group" box.

4. Drag "stimulus" to the "Subgroup" box.

Click OK, and you should get a new table with the data arranged properly.

posted by myeviltwin at 4:59 PM on July 22, 2007

Here's the step-by-step:

1. From the raw data table, choose Table > Summary.

2. Click on RT and under "Statistics" select mean.

3. Drag "Subject" to the "Group" box.

4. Drag "stimulus" to the "Subgroup" box.

Click OK, and you should get a new table with the data arranged properly.

posted by myeviltwin at 4:59 PM on July 22, 2007

Response by poster: myeviltwin: Thanks, your answer lets me do the equivalent of pivottables in jmp.

I still feel like there should be a way to do this without that intermediate step at all, though. I don't really follow what's going on in what a.r.m.o.o.m. pointed to (which is probably a bad sign as far as my understanding of what I'm doing goes) ...

posted by dmd at 5:24 PM on July 22, 2007

I still feel like there should be a way to do this without that intermediate step at all, though. I don't really follow what's going on in what a.r.m.o.o.m. pointed to (which is probably a bad sign as far as my understanding of what I'm doing goes) ...

posted by dmd at 5:24 PM on July 22, 2007

dmd, the ins-and-outs of the many kinds of anovas, why, when, and how they work etc. is a very complicated topic. How to use the last generation of statistical software (which JMP is built off of, but it apparently much friendlier) is also pretty tough. The need for great control and flexibility by professionals lead to a hard learning curve and tough controls. If the long explanation I linked is confusion, that's probably because it's not targeted enough at what you want.

Also, I like AROOM better than my actual handle.

posted by a robot made out of meat at 6:47 PM on July 22, 2007

Also, I like AROOM better than my actual handle.

posted by a robot made out of meat at 6:47 PM on July 22, 2007

I use JMP all the time. For those who aren't familiar, it is made by SAS, but uses a GUI to access SAS functions rather than a command line.

If I understand your question, you're more interested in running ANOVA than in just finding the means, as with the summary function. You can do ANOVA directly from the data you've described with one step in JMP. Under the Analyze menu, choose "Fit Y by X". Make RT your response and both subject and stimulus your factors. On the output screen, you can get it to run Student's t-test and various other types of tests, in addition to ANOVA.

posted by hydropsyche at 9:43 AM on July 23, 2007

If I understand your question, you're more interested in running ANOVA than in just finding the means, as with the summary function. You can do ANOVA directly from the data you've described with one step in JMP. Under the Analyze menu, choose "Fit Y by X". Make RT your response and both subject and stimulus your factors. On the output screen, you can get it to run Student's t-test and various other types of tests, in addition to ANOVA.

posted by hydropsyche at 9:43 AM on July 23, 2007

Response by poster: hydropsyche, I'm pretty sure that's specifically forbidden. All of those individual RT measurements don't 'count' as observations - they are values whose

posted by dmd at 2:27 PM on July 23, 2007

*mean*is an observation.posted by dmd at 2:27 PM on July 23, 2007

Sorry, this is pretty different data from what I do. If that's the case, then I think you're stuck using the summary table and then running ANOVA on that. 2 steps in JMP beats cutting and pasting from Excel.

posted by hydropsyche at 8:42 AM on July 24, 2007

posted by hydropsyche at 8:42 AM on July 24, 2007

This thread is closed to new comments.

posted by a robot made out of meat at 4:00 PM on July 22, 2007