According to SPSS, when calculating a correlationship for ordinal data, select Spearman model; for interval data, use Pearson model. If qdec deal with pain scale as continuous variable, Pearson model should be used. But pain scale is interval data, Spearman model seemly be correct. Dose qdec have statistic method choice? Sorry, too many questions. Thanks. Hong Xie
________________________________
From: Douglas N Greve [mailto:greve@nmr.mgh.harvard.edu] Sent: Mon 1/12/2009 4:03 PM To: Xie, Hong Cc: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] Qdec correlation analyses
why not a continuous model? what model do you want to use?
Xie, Hong wrote:
Thanks a lot. This scrip seemly describes F-test or T-test. In my study, I want to analyze the correlationship between patient's cortical thickness and pain scale (from 0 to 6). After I create data table. Qdec automatically put pain scale into continuous variable category. Actually, pain scale is ordinal variable, not continuous variable. Different correlation model should be used. How should I do it by qdec? I very appreciate your help. Hong Xie
*From:* Douglas N Greve [mailto:greve@nmr.mgh.harvard.edu] *Sent:* Mon 1/12/2009 3:11 PM *To:* Xie, Hong *Cc:* freesurfer@nmr.mgh.harvard.edu *Subject:* Re: [Freesurfer] Qdec correlation analyses
Not with QDEC, but you can write your own FSGD, See
https://surfer.nmr.mgh.harvard.edu/fswiki/FsgdExamples
Xie, Hong wrote:
Hello, Thank you for answering me. I searched mailing list and read this answer email: /I see that group.levels has four levels. Unfortunately, right now qdec only supports two levels for a discrete factor. One the features in the works is to increase this./ /Nick Schmansky Wed, 27 Feb 2008 11:08:10 -0800/ Is this question still here? I have a five levels discrete factor. Can I run correlation statistic by Qdec? Thanks again Hong Xie
*From:* Douglas N Greve [mailto:greve@nmr.mgh.harvard.edu] *Sent:* Thu 1/8/2009 4:40 PM *To:* Xie, Hong *Cc:* freesurfer@nmr.mgh.harvard.edu *Subject:* Re: [Freesurfer] Qdec correlation analyses
It has to be interpreted as a continuous variable. The test is done using a t-test inside the glm. This reduces to a Pearson if there is only one variable and no intercept (which it is not).
doug
Xie, Hong wrote:
Hello, freesurfer group,
I have run qdec for comparing thickness between two groups. When qdec does a correlation analyses for thickness versus a second variable: 1) Does the second variable have to be a continuous interval variable or 2) can the second variable be an ordinal variable with more than 2 levels? In the case of the correlation between two continuous variable, is the test a Pearson correlation? If 2) is possible, is the correlation a Spearman correlation?
Thank for your help,
Hong Xie
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer