proportions as DVs
Locked
 
|
Do any of the more recent spss routines offer a better method for analyzing proportions than what I understand to be the traditional method of applying some type of transformation (arcsine, logit, probit) to the DV to stabilize the variance?
Let me give a specific example: Data: random samples of students in schools (ignore weighting). Student level DV is post high school education attainment scored as five categories. Analysis is at the school level. Thus a DV is proportion of students in a category.
Thanks, Gene Maguin |
|
|
The logit translation is often a good choice.
There is also an extension command named STATS PROPOR REGR that estimates
the proportion equation directly. See the SPSS Community (www.ibm.com/developerworks/spssdevcentral)
Extension commands collection.
Jon Peck (no "h") aka Kim Senior Software Engineer, IBM [hidden email] phone: 720-342-5621 From: "Maguin, Eugene" <[hidden email]> To: [hidden email], Date: 01/21/2014 02:25 PM Subject: [SPSSX-L] proportions as DVs Sent by: "SPSSX(r) Discussion" <[hidden email]> Do any of the more recent spss routines offer a better method for analyzing proportions than what I understand to be the traditional method of applying some type of transformation (arcsine, logit, probit) to the DV to stabilize the variance? Let me give a specific example: Data: random samples of students in schools (ignore weighting). Student level DV is post high school education attainment scored as five categories. Analysis is at the school level. Thus a DV is proportion of students in a category. Thanks, Gene Maguin |
Administrator
|
In reply to this post by Maguin, Eugene
Hi Gene. I suppose your 5 attainment score categories are ordinal, are they?
I know you say analysis is at the school level, but do you have the student level data? If you do, I suspect you could get a suitable model via GENLINMIXED. Here are a couple of examples from the FM: (Nominal) Multinomial logistic regression GENLINMIXED /FIELDS TARGET=bfast /TARGET_OPTIONS DISTRIBUTION=MULTINOMIAL LINK=LOGIT /FIXED EFFECTS=agecat gender active. (Ordinal) Multinomial logistic regression GENLINMIXED /FIELDS TARGET=chist /TARGET_OPTIONS DISTRIBUTION=MULTINOMIAL LINK=CLOGLOG /FIXED EFFECTS=numcred othnstal housng age duration. Of course, you could add school-level explanatory variables and random effects, etc. HTH.
--
Bruce Weaver bweaver@lakeheadu.ca http://sites.google.com/a/lakeheadu.ca/bweaver/ "When all else fails, RTFM." PLEASE NOTE THE FOLLOWING: 1. My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above. 2. The SPSSX Discussion forum on Nabble is no longer linked to the SPSSX-L listserv administered by UGA (https://listserv.uga.edu/). |
|
|
Yes, they are ordinal but I don't know that they would meet the proportional odds assumption for ordinal regression. I'm looking at the Genlinmixed documentation now but I don't see anything to indicate that that could be tested as it can in Plum. We do have student level data but the analysis is to use school level variables such as free lunch percentage, percent minority, teacher FTE per student, etc. The school level IVs are not aggregated student variables. Thus it's really a multilevel analysis and my impression, quite possibly wrong, is that Genlinmixed can't do a multilevel analysis comparable to Mixed.
Jon (thank you) suggested a logit transform and we can do that in spss. I'm not so sure about this but I also think that we can use GLM as a multivariate regression routine. Gene -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Bruce Weaver Sent: Tuesday, January 21, 2014 5:12 PM To: [hidden email] Subject: Re: proportions as DVs Hi Gene. I suppose your 5 attainment score categories are ordinal, are they? I know you say analysis is at the school level, but do you have the student level data? If you do, I suspect you could get a suitable model via GENLINMIXED. Here are a couple of examples from the FM: *(Nominal) Multinomial logistic regression* GENLINMIXED /FIELDS TARGET=bfast /TARGET_OPTIONS DISTRIBUTION=MULTINOMIAL LINK=LOGIT /FIXED EFFECTS=agecat gender active. *(Ordinal) Multinomial logistic regression* GENLINMIXED /FIELDS TARGET=chist /TARGET_OPTIONS DISTRIBUTION=MULTINOMIAL LINK=CLOGLOG /FIXED EFFECTS=numcred othnstal housng age duration. Of course, you could add school-level explanatory variables and random effects, etc. HTH. Maguin, Eugene wrote > Do any of the more recent spss routines offer a better method for > analyzing proportions than what I understand to be the traditional > method of applying some type of transformation (arcsine, logit, > probit) to the DV to stabilize the variance? Let me give a specific > example: Data: random samples of students in schools (ignore > weighting). Student level DV is post high school education attainment > scored as five categories. Analysis is at the school level. Thus a DV is proportion of students in a category. > > Thanks, Gene Maguin ----- -- Bruce Weaver [hidden email] http://sites.google.com/a/lakeheadu.ca/bweaver/ "When all else fails, RTFM." NOTE: My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above. -- View this message in context: http://spssx-discussion.1045642.n5.nabble.com/proportions-as-DVs-tp5724051p5724053.html Sent from the SPSSX Discussion mailing list archive at Nabble.com. ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |