search

IALSA / IALSA-2015-Portland

Hub to accompany IALSA 2015 workshop at Portland, OR, Feb 22-25, 2015
GNU General Public License v2.0
2 stars 0 forks source link

PxP - Covariates and Centering #73

Open andkov opened 9 years ago

andkov commented 9 years ago

Please use the thread to arrive at 1) Common names of covariates used in MPlus model estimation. 2) Rules for centering of the covariates

andkov commented 9 years ago

males and females have different centers

andkov commented 9 years ago

@ampiccinin, @GracielaMuniz , @annierobi , i'm informing you that currently there is discord among the models with respect to the age centering. I observed the following:

HRS - centered at 75
LASA - centered at 80 
OCTO - centered at 80
RADC - centered at 70

Please develop the centering convention, and post your decision here so we may have it for future reference.

annierobi commented 9 years ago

Actually, I think we had discussed centering at age 70 except for Octo-Twin which is centered at 80 because it's a sample of older old. I must have forgotten that rule after running the octo models and ran LASA with 80. I will center at age 70. As for HRS, there must be a reason for using 75 but I don't know what it is.

On Thu, Oct 22, 2015 at 10:36 PM, Andriy V. Koval notifications@github.com wrote:

@ampiccinin https://github.com/ampiccinin, @GracielaMuniz https://github.com/GracielaMuniz , @annierobi https://github.com/annierobi , i'm informing you that currently there is discord among the models with respect to the age centering. I observed the following:

HRS - centered at 75 LASA - centered at 80 OCTO - centered at 80 RADC - centered at 70

Please develop the centering convention, and post your decision here so we may have it for future reference.

— Reply to this email directly or view it on GitHub https://github.com/IALSA/IALSA-2015-Portland/issues/73#issuecomment-150416401 .

chenkaiwu commented 9 years ago

If I remember correctly, the mean age of the HRS sample at baseline was around 75. Andrey, do you think it is necessary to be consistent on age centering?

From: annierobi [mailto:notifications@github.com] Sent: Thursday, October 22, 2015 7:51 PM To: IALSA/IALSA-2015-Portland IALSA-2015-Portland@noreply.github.com Subject: Re: [IALSA-2015-Portland] PxP - Covariates and Centering (#73)

Actually, I think we had discussed centering at age 70 except for Octo-Twin which is centered at 80 because it's a sample of older old. I must have forgotten that rule after running the octo models and ran LASA with 80. I will center at age 70. As for HRS, there must be a reason for using 75 but I don't know what it is.

On Thu, Oct 22, 2015 at 10:36 PM, Andriy V. Koval <notifications@github.com mailto:notifications@github.com > wrote:

@ampiccinin https://github.com/ampiccinin, @GracielaMuniz https://github.com/GracielaMuniz , @annierobi https://github.com/annierobi , i'm informing you that currently there is discord among the models with respect to the age centering. I observed the following:

HRS - centered at 75 LASA - centered at 80 OCTO - centered at 80 RADC - centered at 70

Please develop the centering convention, and post your decision here so we may have it for future reference.

— Reply to this email directly or view it on GitHub https://github.com/IALSA/IALSA-2015-Portland/issues/73#issuecomment-150416401 .

— Reply to this email directly or view it on GitHub https://github.com/IALSA/IALSA-2015-Portland/issues/73#issuecomment-150425005 . https://github.com/notifications/beacon/AKkM1TnK_DI5SE96a5sKe-_luBH15Nxzks5o-ZgTgaJpZM4FtvM8.gif

andkov commented 9 years ago

@chenckaiwu, I think so. As far as can understand it is important for cross-study comparison

chenkaiwu commented 9 years ago

Sure, so are we all using 70?

ampiccinin commented 9 years ago

Let me double-check age centring across all studies in the morning. Based on our focus and assumptions for the PxP project, some variation in age centring should have little impact. I am reluctant to ask too many people to re-run models too many times!!

If we DO re-run anything, we should add the save data fscores syntax and make slope-slope scatter plots at the same time.

I'm aiming to circulate a first draft in the next day or so as well.

Thanks! Best, Andrea

On Oct 22, 2015, at 7:51 PM, annierobi notifications@github.com<mailto:notifications@github.com> wrote:

Actually, I think we had discussed centering at age 70 except for Octo-Twin which is centered at 80 because it's a sample of older old. I must have forgotten that rule after running the octo models and ran LASA with 80. I will center at age 70. As for HRS, there must be a reason for using 75 but I don't know what it is.

On Thu, Oct 22, 2015 at 10:36 PM, Andriy V. Koval notifications@github.com<mailto:notifications@github.com> wrote:

@ampiccinin https://github.com/ampiccinin, @GracielaMuniz https://github.com/GracielaMuniz , @annierobi https://github.com/annierobi , i'm informing you that currently there is discord among the models with respect to the age centering. I observed the following:

HRS - centered at 75 LASA - centered at 80 OCTO - centered at 80 RADC - centered at 70

Please develop the centering convention, and post your decision here so we may have it for future reference.

Reply to this email directly or view it on GitHub https://github.com/IALSA/IALSA-2015-Portland/issues/73#issuecomment-150416401 .

Reply to this email directly or view it on GitHubhttps://github.com/IALSA/IALSA-2015-Portland/issues/73#issuecomment-150425005.

ampiccinin commented 9 years ago

Hi all,

Given that we are focusing on the random effects, and are estimating linear change, the small differences in age centering for a couple of studies should have a negligible impact on the results. It will impact the fixed effects, but we are not currently worrying about those. Since not all studies have an age of 70 within the range of their data, we can't center all at 70 anyway. Similarly for education two studies are having to resort to a dichotomous variable.