On 10/27/2021 11:39 AM, Douglas N. Greve wrote:
Dealing with discrete variables as discrete is the way I would do it as you would not have any reason to think that there is some linear relationship between your the (arbitrary) values you assign to discrete variables and thickness. Age is not an arbitrary value, so I would use that as a continuous variable.
On 10/27/2021 6:06 AM, Thalhammer, Melissa wrote:
External Email - Use Caution
Dear FreeSurfer users,
I am interested in cortical thickness differences between preterm (PT) and fullterm (FT) individuals. I have used DODS with mri_glmfit and wanted to regress out scanner (4 sites) and sex in the analysis, using the attached FSGD file. Both of the variables are discrete and I have found different information on how to include discrete variables into the model.
I have followed Andrew Jahn's blog at first who includes discrete variables in the variable list just as continuous ones and specifies them with values of 0 or 1 (*MailScanner has detected a possible fraud attempt from "secure-web.cisco.com" claiming to be* https://andysbrainbook.readthedocs.io/en/latest/FreeSurfer/FS_ShortCourse/FS... https://secure-web.cisco.com/1mCX947TtGflWAdnM3CtHlWXx5Aa4vkeu9B25dEdIVjqujqkljqo76fFj56AFCqedaxfFDqlldd8ejrek89Cr6-IDZfyFpPLZNnSfAgx554zrtzB8GXnwzUBpv1C5tcq17LGjh0UZVCUAytdtAHJzahIBfCoxklF9YvKFJsvkdOn2a1IthQ8R1YRR-YvR4jWe60jAJ_rK1Y-g1ZxoH3BGr0U2-ks39W954G6KAX4BV7RuebOy9Jp2gs-5LSXtivyNoB3rEB7aA1_b09hyqrwKXQ/https%3A%2F%2Fandysbrainbook.readthedocs.io%2Fen%2Flatest%2FFreeSurfer%2FFS_ShortCourse%2FFS_07_FSGD.html). I have adapted this procedure and have specified my gender values as 1 and 2 (simply because it was stated like this in the participants information sheet I have received).
Recently, I have read a post on the mail archive, where Doug states: "you should not code gender and handedness as a continuous variables. In doing this, you are saying that you expect right handers to have 2x the thickness as left handers (or females to have 2x the thickness as males)" (*MailScanner has detected a possible fraud attempt from "secure-web.cisco.com" claiming to be* https://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg27832.html https://secure-web.cisco.com/1A4RpPVoEFVSno401uhBm0BawiKw27S9-5VpEg7YyqN3rIyrimpbhqGdB5I3dtTasdY4_o1uitlMHH_nc7Wyblr3HpGh3qfU7YCP7RZMK4u_hLFEtfAgLNTvdMaoPRoCGO7BAms6XaUKXHOPXEVlq26kO_QxHOx-Xeh4A8gcNKr5gnaKqM9PN0n1I9Ohy3nqyXZ7dOl0Hc3fgxKfVe9wXlhxkEYC7iMMdTxnRBwpXmGx2ejdNwMhusroh0TDrvKcJ7eZ_bD0XIPvOK_ZSbM8XxA/https%3A%2F%2Fwww.mail-archive.com%2Ffreesurfer%40nmr.mgh.harvard.edu%2Fmsg27832.html). When following this argument, this would also hold for continuous variables, i.e., a patient aged 50 would have 2x the thickness of a patient aged 25.
So my question is how to correctly deal with discrete variables and why you have suggested to code gender as classes instead of variables. What is the actual difference? How does the model treat classes and variables?
Cheers and many thanks in advance,
Melissa Thalhammer
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer