FW: Multiple imputation
Locked
 
|
Dear people on the list,
Someone from the list (Jan Sonsmeyer) responded to a message that I posted on the SPSS mailing list a while ago. He wanted to combine the results of ANOVA of multiply imputed data sets. In my message, I wrote that the possibilities of doing this were still quite limited in PASW. However, rules for combining F-tests in, for example, ANOVA, do exist (See, for example, Rubin, 1987; Schafer, 1997). They can be applied using a PASW macro that I've programmed. I recommended this macro to Mr. Sonsmeyer and I would like to recommend this procedure to anyone who wants to combine the results of ANOVA (or any F-test) of multiply imputed data sets. It can be downloaded from the link in my response to Mr. Sonsmeyer (see below). However, wouldn't it be better if SPSS were to implement this in future PASW releases? After all, it is an existing procedure and there is obviously a need among PASW users to combine the results of ANOVA of multiply imputed data sets.
Best regards,
Joost van Ginkel
Van: Ginkel, Joost van Verzonden: wo 7-10-2009 16:44 Aan: 'Jan Sohnsmeyer' Onderwerp: RE: Multiple imputation Dear Mr. Sonsmeyer,
Did you ask for parameter estimates in the output? The multiple imputation procedure in PASW only combines parameter estimates and their standard errors. If you would like to combine the overall F-tests of each effect, you can use my PASW macro for combining the results of multiple imputation for multidimensional estimates. They can be downloaded from: http://www.datatheory.nl/pages/ginkel.html (mi-mul.zip). A manual is also included. Unfortunately, this is much more time-consuming than combining the results the regular way, but you will end up with overall F-tests of each effect. I'm actually surprised that SPSS has not included this combination procedure in PASW (yet). By the way, is it ok if I post your message and my response to it on the mailing list? It may encourage SPSS to work on this in future releases. Until then, people with the same problem can use my macro. Well, thank you very much in advance.
Best regards,
Joost van Ginkel
Joost R. Van Ginkel, PhD From: Jan Sohnsmeyer [mailto:[hidden email]] Sent: 07 October 2009 16:28 To: Ginkel, Joost van Subject: Multiple imputation Dear Mr. van Ginkel,
I am writing to ask for help concerning some problems with pooling data after multiple Imputation in SPSS 17. After running my required repeated measures ANOVA my output shows the results of each imputation, but not the pooled results. “Harry” posted the same question to the SPSSX-L mailing list and you posted the following response:
--- Response in SPSSX-L archives --- Dear Harry, To date, no combination rules for MI in (repeated measures) ANOVA or MANOVA have been defined. For ANOVA you may resolve the problem by Using a multilevel model in Mixed models, where repeated measures are nested within subjects. SPSS can give pooled results within Mixed models. For MANOVA there is not really a solution at hand yet. Best regards, Joost van Ginkel ---
I tried out your suggestion but was not able to get the pooled results within mixed models. Maybe you can give me a more detailed description of what to do.
I am looking forward to hear from you.
Sincerely, Jan Sohnsmeyer
-------------------------------------------------------------------- Jan Sohnsmeyer Wissenschaftlicher Mitarbeiter Institut für Sport und Sportwissenschaften der CAU zu Kiel Olshausenstr. 74, 24098 Kiel fon: +49 431 880 3767 fax: +49 431 880 3773 mail: [hidden email]
|
|
|
Strange, you shouldn't need a password, nor the installation of some other programs. I've tried to open the link but it didn't work either, although I didn't have to enter any password. Perhaps you can copy the link and paste it in internet explorer. If that fails, try www.datatheory.nl. Then click on faculty and staff, go to my personal page (Joost van Ginkel) and there you'll find it.
Best regards,
Joost van Ginkel Van: Peck, Jon [mailto:[hidden email]] Verzonden: wo 7-10-2009 23:00 Aan: Ginkel, Joost van Onderwerp: RE: [SPSSX-L] FW: Multiple imputation The link you posted requires a username and password and the installation and execution of some other programs. Can you make this easier?
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Ginkel, Joost van
Dear people on the list,
Someone from the list (Jan Sonsmeyer) responded to a message that I posted on the SPSS mailing list a while ago. He wanted to combine the results of ANOVA of multiply imputed data sets. In my message, I wrote that the possibilities of doing this were still quite limited in PASW. However, rules for combining F-tests in, for example, ANOVA, do exist (See, for example, Rubin, 1987; Schafer, 1997). They can be applied using a PASW macro that I've programmed. I recommended this macro to Mr. Sonsmeyer and I would like to recommend this procedure to anyone who wants to combine the results of ANOVA (or any F-test) of multiply imputed data sets. It can be downloaded from the link in my response to Mr. Sonsmeyer (see below). However, wouldn't it be better if SPSS were to implement this in future PASW releases? After all, it is an existing procedure and there is obviously a need among PASW users to combine the results of ANOVA of multiply imputed data sets.
Best regards,
Joost van Ginkel
Van: Ginkel, Joost van Dear Mr. Sonsmeyer,
Did you ask for parameter estimates in the output? The multiple imputation procedure in PASW only combines parameter estimates and their standard errors. If you would like to combine the overall F-tests of each effect, you can use my PASW macro for combining the results of multiple imputation for multidimensional estimates. They can be downloaded from: http://www.datatheory.nl/pages/ginkel.html (mi-mul.zip). A manual is also included. Unfortunately, this is much more time-consuming than combining the results the regular way, but you will end up with overall F-tests of each effect. I'm actually surprised that SPSS has not included this combination procedure in PASW (yet). By the way, is it ok if I post your message and my response to it on the mailing list? It may encourage SPSS to work on this in future releases. Until then, people with the same problem can use my macro. Well, thank you very much in advance.
Best regards,
Joost van Ginkel Joost R. Van Ginkel, PhD
From: Jan Sohnsmeyer [mailto:[hidden email]] Dear Mr. van Ginkel,
I am writing to ask for help concerning some problems with pooling data after multiple Imputation in SPSS 17. After running my required repeated measures ANOVA my output shows the results of each imputation, but not the pooled results. “Harry” posted the same question to the SPSSX-L mailing list and you posted the following response:
--- Response in SPSSX-L archives --- Dear Harry, To date, no combination rules for MI in (repeated measures) ANOVA or MANOVA have been defined. For ANOVA you may resolve the problem by Using a multilevel model in Mixed models, where repeated measures are nested within subjects. SPSS can give pooled results within Mixed models. For MANOVA there is not really a solution at hand yet. Best regards, Joost van Ginkel ---
I tried out your suggestion but was not able to get the pooled results within mixed models. Maybe you can give me a more detailed description of what to do.
I am looking forward to hear from you.
Sincerely, Jan Sohnsmeyer
-------------------------------------------------------------------- Jan Sohnsmeyer Wissenschaftlicher Mitarbeiter Institut für Sport und Sportwissenschaften der CAU zu Kiel Olshausenstr. 74, 24098 Kiel fon: +49 431 880 3767 fax: +49 431 880 3773 mail: [hidden email]
********************************************************************** This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error please notify the system manager. ********************************************************************** |
Administrator
|
Hi Joost. This is very timely for me, as I've just begun exploring MI in SPSS 17. But neither of the methods described above is working for me. I ultimately end up on http://www.leidenuniv.nl/. Regarding one of the e-mail messages you included in your earlier post, you suggested that if one uses MIXED, one does get pooled F-tests, is that right? So why not just use MIXED to run the ANOVA model? Or am I missing something? Cheers, Bruce
--
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/). |
|
|
In reply to this post by Joost van Ginkel
Hi Everyone,
Can someone explain to me why I get the following data in PASW 17: 1 0 1 2 . a a 1 a v 2 a 1 . 1 when I import an Excel file containing this information which doesn't have a column header. (note values for obs2 and obs5 in column 2, the problematic column, are blanks). 1 a 1 2 a a 1 a v 2 a 1 1 However, if I put a column header in Excel, SPSS correctly imports it and shows this: (same as above) 1 a 1 2 a a 1 a v 2 a 1 1 Thanks. Florio ===================== 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 |
Re: FW: Multiple imputation
|
|
In reply to this post by Bruce Weaver
<[hidden email]>
A<[hidden email]> From: "Ginkel, Joost van" <[hidden email]> To: "Bruce Weaver" <[hidden email]>, <[hidden email]> X-Scanned-By: Digested by UGA Mail Gateway on 128.192.1.75 Content-Transfer-Encoding: 8bit X-MIME-Autoconverted: from quoted-printable to 8bit by listserv.uga.edu id n988Er9a031908 Dear Bruce and John, First of all, the macros can also be downloaded from http://www.uvt.nl/faculteiten/fsw/organisatie/departementen/mto/software 2.html. If that still doesn't work, I can always send the files by email to anyone who is interested (just send me an email with a request). In the mean time I will try to figure out why several people are experiencing trouble downloading them. As for the F-tests in mixed models, my original message didn't say that pooled F-tests were provided in the output. My message was a response to someone who wanted to carry out a repeated measures ANOVA in PASW for a multipy imputed data set. I said that he should try mixed models instead with repeated measures being nested within subjects. Unfortunately, this will not provide any F-tests either. However, I wasn't totally right about the unavailability of combination rules for multiple imputation in ANOVA as they do exist, and can be applied using the macros. Best regards, Joost van Ginkel Joost R. Van Ginkel, PhD Leiden University Faculty of Social and Behavioural Sciences Data Theory Group PO Box 9555 2300 RB Leiden The Netherlands Tel: +31-(0)71-527 3620 Fax: +31-(0)71-527 1721 -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Bruce Weaver Sent: 08 October 2009 00:14 To: [hidden email] Subject: Re: FW: Multiple imputation Ginkel, Joost van wrote: > > Strange, you shouldn't need a password, nor the installation of some > other programs. I've tried to open the link but it didn't work either, > although I didn't have to enter any password. Perhaps you can copy the > link and paste it in internet explorer. If that fails, try > www.datatheory.nl <http://www.datatheory.nl/> . Then click on faculty > and staff, go to my personal page (Joost van Ginkel) and there you'll find it. > > Best regards, > > Joost van Ginkel > Hi Joost. This is very timely for me, as I've just begun exploring MI in SPSS 17. But neither of the methods described above is working for me. I ultimately end up on http://www.leidenuniv.nl/. Regarding one of the e-mail messages you included in your earlier post, you suggested that if one uses MIXED, one does get pooled F-tests, is that right? So why not just use MIXED to run the ANOVA model? Or am I missing something? Cheers, Bruce ----- -- 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://www.nabble.com/FW%3A-Multiple-imputation-tp25794139p25795315.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 ********************************************************************** This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error please notify the system manager. ********************************************************************** ===================== 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 |
Re: FW: Multiple imputation
|
|
In reply to this post by Joost van Ginkel
Dear Bruce and John, First of all, the macros can also be downloaded from http://www.uvt.nl/faculteiten/fsw/organisatie/departementen/mto/software2.html. If that still doesn't work, I can always send the files by email to anyone who is interested (just send me an email with a request). In the mean time I will try to figure out why several people are experiencing trouble downloading them. As for the F-tests in mixed models, my original message didn't say that pooled F-tests were provided in the output. My message was a response to someone who wanted to carry out a repeated measures ANOVA in PASW for a multipy imputed data set. I said that he should try mixed models instead with repeated measures being nested within subjects. Unfortunately, this will not provide any F-tests either. However, I wasn't totally right about the unavailability of combination rules for multiple imputation in ANOVA as they do exist, and can be applied using the macros. Best regards, Joost van Ginkel Van: SPSSX(r) Discussion namens Bruce Weaver Verzonden: do 8-10-2009 0:13 Aan: [hidden email] Onderwerp: Re: FW: Multiple imputation Ginkel, Joost van wrote: |
|
Administrator
|
In reply to this post by Undetermined origin c/o LISTSERV administrator
Thanks Joost. After fixing a line-wrap problem, I was able to get the file this time.
--
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/). |
|
|
In reply to this post by Joost van Ginkel
Although this is a very old post, there is quite an important update on the subject of multiple imputation and analysis of variance. Recently a paper of mine got published which describes the procedure for combining the results of (repeated measures) ANOVA on multiply imputed data sets, both from a theoretical and practical point of view. More specifically, the paper describes how the results of ANOVA can be combined in several software packages, including SPSS, using the SPSS macro mentioned in the earlier replies to this post. The reference for this paper is:
Van Ginkel, J.R. & Kroonenberg, P.M. (2014). Analysis of variance of multiply imputed data. Multivariate Behavioral Research, 49, 78-91. Earlier I posted another message on this subject but I wanted to use as many ways as possible to make people aware of this important topic, including replying to earlier posts on this. Hence this reply. Best, Joost |
Administrator
|
In reply to this post by Joost van Ginkel
Hi Joost,
Clicking the link in the response redirects me to a login page. However copy/paste gets me to your page. http://www.datatheory.nl/pages/ginkel.html I agree that this should be in the lap of IBM. Unfortunately it seems to me that new developments in statistical procedure development have been relegated to R and python scripters. I can't think of any significant new INTERNAL statistical innovations since GENLINMIXED. ver 21/22 certainly hasn't added anything revolutionary ... snooze. I would be interesting to catalog the revolutionary innovations since say version 11.5 . Not much AFAICT.
Please reply to the list and not to my personal email.
Those desiring my consulting or training services please feel free to email me. --- "Nolite dare sanctum canibus neque mittatis margaritas vestras ante porcos ne forte conculcent eas pedibus suis." Cum es damnatorum possederunt porcos iens ut salire off sanguinum cliff in abyssum?" |
|
Administrator
|
DOH: I just realized I replied to a post from 2009 ;-)
My comment still applies! --
Please reply to the list and not to my personal email.
Those desiring my consulting or training services please feel free to email me. --- "Nolite dare sanctum canibus neque mittatis margaritas vestras ante porcos ne forte conculcent eas pedibus suis." Cum es damnatorum possederunt porcos iens ut salire off sanguinum cliff in abyssum?" |
Release History [was Re: FW: Multiple imputation]
|
|
In reply to this post by David Marso
There are a couple of ways to get at the
release history:
1. There's a list, by syntax command, at http://pic.dhe.ibm.com/infocenter/spssstat/v22r0m0/topic/com.ibm.spss.statistics.reference/spss/base/syn_refintro_history.htm 2. The Overview topic lists all the "What's New?" topics, by release, at http://pic.dhe.ibm.com/infocenter/spssstat/v22r0m0/topic/com.ibm.spss.statistics.help/spss/base/overvw_container.htm Alex From: David Marso <[hidden email]> To: [hidden email], Date: 03/13/2014 12:25 PM Subject: Re: FW: Multiple imputation Sent by: "SPSSX(r) Discussion" <[hidden email]> Hi Joost, Clicking the link in the response redirects me to a login page. However copy/paste gets me to your page. http://www.datatheory.nl/pages/ginkel.html I agree that this should be in the lap of IBM. Unfortunately it seems to me that new developments in statistical procedure development have been relegated to R and python scripters. I can't think of any significant new INTERNAL statistical innovations since GENLINMIXED. ver 21/22 certainly hasn't added anything revolutionary ... snooze. I would be interesting to catalog the revolutionary innovations since say version 11.5 . Not much AFAICT. |
Re: Release History [was Re: FW: Multiple imputation]
|
Administrator
|
And we conclude from this?
Drum roll....glacial progress towards a statistical revolution ;-)
Please reply to the list and not to my personal email.
Those desiring my consulting or training services please feel free to email me. --- "Nolite dare sanctum canibus neque mittatis margaritas vestras ante porcos ne forte conculcent eas pedibus suis." Cum es damnatorum possederunt porcos iens ut salire off sanguinum cliff in abyssum?" |
| Free forum by Nabble | Edit this page |