multiple imputation:pooled data
Locked
multiple imputation:pooled data
 
|
Hi
I'm doing a multiple imputation for missing data. i have to use these data for HLM later. I need the pooled dataset in order to use it instead that the different single imputations. How can I obtain the pooled dataset? THANK YOU VERY MUCH maria |
Re: multiple imputation:pooled data
Administrator
|
What is a "pooled dataset"? The way multiple imputation typically works is that you run your model multiple times, once per imputed data set, and then work out pooled estimates of the model parameters and standard errors (using methods described by Donald Rubin). That same basic approach is used for HLM, I should think. Do you know something that suggests otherwise?
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/). |
Re: multiple imputation:pooled data
|
|
I saw this before, and was uncertain myself, since SAS and SPSS use stacked files for MLM/HLM analysis. I also don't believe that either let you run an mixed model analysis on multiply imputed data (I could be wrong, and would love to know otherwise).
If the requester uses HLM specifically, that uses separate MDM data files for each level of data. Doing a multiple imputation with this data requires that you run a separate multiple imputation on each MDM file for each level (for the sake of argument, you could do this with SPSS). You save each of these as a separate files, and import them each as separate MDM files for a given level. HLM allows you to run a multiple imputation analysis for your HLM from this, but the actual creation of the 5 or so sets of data for each level must be done externally. Matthew J Poes Research Data Specialist Center for Prevention Research and Development University of Illinois 510 Devonshire Dr. Champaign, IL 61820 Phone: 217-265-4576 email: [hidden email] -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Bruce Weaver Sent: Thursday, March 15, 2012 2:23 PM To: [hidden email] Subject: Re: multiple imputation:pooled data What is a "pooled dataset"? The way multiple imputation typically works is that you run your model multiple times, once per imputed data set, and then work out pooled estimates of the model parameters and standard errors (using methods described by Donald Rubin). That same basic approach is used for HLM, I should think. Do you know something that suggests otherwise? HTH. mariaspinelli wrote > > Hi > > I'm doing a multiple imputation for missing data. i have to use these > data for HLM later. I need the pooled dataset in order to use it > instead that the different single imputations. How can I obtain the pooled dataset? > > THANK YOU VERY MUCH > > > > maria > ----- -- 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/multiple-imputation-pooled-data-tp5568656p5569041.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 |
Re: multiple imputation:pooled data
|
|
SAS will certainly do it using the by replicate subcommand. I think SPSS has a similar procedure.
Dr. Paul R. Swank, Children's Learning Institute Professor, Department of Pediatrics, Medical School Adjunct Professor, School of Public Health University of Texas Health Science Center-Houston -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Poes, Matthew Joseph Sent: Thursday, March 15, 2012 2:31 PM To: [hidden email] Subject: Re: multiple imputation:pooled data I saw this before, and was uncertain myself, since SAS and SPSS use stacked files for MLM/HLM analysis. I also don't believe that either let you run an mixed model analysis on multiply imputed data (I could be wrong, and would love to know otherwise). If the requester uses HLM specifically, that uses separate MDM data files for each level of data. Doing a multiple imputation with this data requires that you run a separate multiple imputation on each MDM file for each level (for the sake of argument, you could do this with SPSS). You save each of these as a separate files, and import them each as separate MDM files for a given level. HLM allows you to run a multiple imputation analysis for your HLM from this, but the actual creation of the 5 or so sets of data for each level must be done externally. Matthew J Poes Research Data Specialist Center for Prevention Research and Development University of Illinois 510 Devonshire Dr. Champaign, IL 61820 Phone: 217-265-4576 email: [hidden email] -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Bruce Weaver Sent: Thursday, March 15, 2012 2:23 PM To: [hidden email] Subject: Re: multiple imputation:pooled data What is a "pooled dataset"? The way multiple imputation typically works is that you run your model multiple times, once per imputed data set, and then work out pooled estimates of the model parameters and standard errors (using methods described by Donald Rubin). That same basic approach is used for HLM, I should think. Do you know something that suggests otherwise? HTH. mariaspinelli wrote > > Hi > > I'm doing a multiple imputation for missing data. i have to use these > data for HLM later. I need the pooled dataset in order to use it > instead that the different single imputations. How can I obtain the pooled dataset? > > THANK YOU VERY MUCH > > > > maria > ----- -- 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/multiple-imputation-pooled-data-tp5568656p5569041.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 ===================== 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: multiple imputation:pooled data
|
|
Thanks Paul, that's really good to know. I learn new things every day on here. I've been utilizing HLM for anything like this in the past. I'll have to review the manual and look at the "by replicate" subcommand to see how this is done.
Matthew J Poes Research Data Specialist Center for Prevention Research and Development University of Illinois 510 Devonshire Dr. Champaign, IL 61820 Phone: 217-265-4576 email: [hidden email] -----Original Message----- From: Swank, Paul R [mailto:[hidden email]] Sent: Thursday, March 15, 2012 3:22 PM To: Poes, Matthew Joseph; [hidden email] Subject: RE: multiple imputation:pooled data SAS will certainly do it using the by replicate subcommand. I think SPSS has a similar procedure. Dr. Paul R. Swank, Children's Learning Institute Professor, Department of Pediatrics, Medical School Adjunct Professor, School of Public Health University of Texas Health Science Center-Houston -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Poes, Matthew Joseph Sent: Thursday, March 15, 2012 2:31 PM To: [hidden email] Subject: Re: multiple imputation:pooled data I saw this before, and was uncertain myself, since SAS and SPSS use stacked files for MLM/HLM analysis. I also don't believe that either let you run an mixed model analysis on multiply imputed data (I could be wrong, and would love to know otherwise). If the requester uses HLM specifically, that uses separate MDM data files for each level of data. Doing a multiple imputation with this data requires that you run a separate multiple imputation on each MDM file for each level (for the sake of argument, you could do this with SPSS). You save each of these as a separate files, and import them each as separate MDM files for a given level. HLM allows you to run a multiple imputation analysis for your HLM from this, but the actual creation of the 5 or so sets of data for each level must be done externally. Matthew J Poes Research Data Specialist Center for Prevention Research and Development University of Illinois 510 Devonshire Dr. Champaign, IL 61820 Phone: 217-265-4576 email: [hidden email] -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Bruce Weaver Sent: Thursday, March 15, 2012 2:23 PM To: [hidden email] Subject: Re: multiple imputation:pooled data What is a "pooled dataset"? The way multiple imputation typically works is that you run your model multiple times, once per imputed data set, and then work out pooled estimates of the model parameters and standard errors (using methods described by Donald Rubin). That same basic approach is used for HLM, I should think. Do you know something that suggests otherwise? HTH. mariaspinelli wrote > > Hi > > I'm doing a multiple imputation for missing data. i have to use these > data for HLM later. I need the pooled dataset in order to use it > instead that the different single imputations. How can I obtain the pooled dataset? > > THANK YOU VERY MUCH > > > > maria > ----- -- 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/multiple-imputation-pooled-data-tp5568656p5569041.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 ===================== 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: multiple imputation:pooled data
|
|
In reply to this post by Swank, Paul R
Thank all.
My advisor ask me to make the MDM file for the HLM program ready without missing datas. She asked me to make the Multiple imputation data and take the pooled dataset in order to use it for the HLM7 program. That's why I was wondering to know if there is a way to get a pooled dataset. With pooled dataset I intend a dataset without missing data who is the summary of the different imputation datasets... Could you give me more indications about the replicate subcommand of SPSS? Thanks you all Maria Il 15/03/2012 15:33, Swank, Paul R [via SPSSX Discussion] ha scritto: SAS will certainly do it using the by replicate subcommand. I think SPSS has a similar procedure. |
Re: multiple imputation:pooled data
|
Administrator
|
In reply to this post by Poes, Matthew Joseph-2
Matthew, "linear mixed models" are included in the list of procedures that support pooling.
Linear Mixed Models • The Descriptive Statistics table supports Mean and N at Naïve pooling. • The Estimates of Fixed Effects table supports Estimate at Univariate pooling. • The Estimates of Covariance Parameters table supports Estimate at Univariate pooling. • The Estimated Marginal Means: Estimates table supports Mean at Univariate pooling. • The Estimated Marginal Means: Pairwise Comparisons table supports Mean Difference at Univariate pooling. From: http://publib.boulder.ibm.com/infocenter/spssstat/v20r0m0/index.jsp?topic=%2Fcom.ibm.spss.statistics.help%2Fmi_analysis.htm 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/). |
Re: multiple imputation:pooled data
|
|
Thank's Bruce, I had apparently missed that. I've only been using the Linear Mixed models in SPSS for about a year now, and was not as familiar. I had taken some time to read up on it in the IBM package before, but apparently missed that it could be used with the MI as well. This is good to know.
I've only had to perform a multiple imputation on data and then perform an MLM once, and this was done with HLM, so that is really the only software I have direct experience using in this regard. I had been told at the time it was one of the only that could do this, and apparently never updated my knowledge on the truth of that situation. Many thanks, as this was something I was hoping was possible or would be possible in future versions. Matthew J Poes Research Data Specialist Center for Prevention Research and Development University of Illinois 510 Devonshire Dr. Champaign, IL 61820 Phone:� 217-265-4576 email: [hidden email] -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Bruce Weaver Sent: Friday, March 16, 2012 10:28 AM To: [hidden email] Subject: Re: multiple imputation:pooled data Matthew, "linear mixed models" are included in the list of procedures that support pooling. Linear Mixed Models • The Descriptive Statistics table supports Mean and N at Naïve pooling. • The Estimates of Fixed Effects table supports Estimate at Univariate pooling. • The Estimates of Covariance Parameters table supports Estimate at Univariate pooling. • The Estimated Marginal Means: Estimates table supports Mean at Univariate pooling. • The Estimated Marginal Means: Pairwise Comparisons table supports Mean Difference at Univariate pooling. From: http://publib.boulder.ibm.com/infocenter/spssstat/v20r0m0/index.jsp?topic=%2Fcom.ibm.spss.statistics.help%2Fmi_analysis.htm Cheers, Bruce Matthew J Poes wrote > > I saw this before, and was uncertain myself, since SAS and SPSS use > stacked files for MLM/HLM analysis. I also don't believe that either > let you run an mixed model analysis on multiply imputed data (I could > be wrong, and would love to know otherwise). > > If the requester uses HLM specifically, that uses separate MDM data > files for each level of data. Doing a multiple imputation with this > data requires that you run a separate multiple imputation on each MDM > file for each level (for the sake of argument, you could do this with > SPSS). You save each of these as a separate files, and import them > each as separate MDM files for a given level. HLM allows you to run a > multiple imputation analysis for your HLM from this, but the actual > creation of the 5 or so sets of data for each level must be done externally. > > Matthew J Poes > Research Data Specialist > Center for Prevention Research and Development University of Illinois > 510 Devonshire Dr. > Champaign, IL 61820 > Phone: 217-265-4576 > email: mpoes@ > > > -----Original Message----- > From: SPSSX(r) Discussion [mailto:SPSSX-L@.UGA] On Behalf Of Bruce Weaver > Sent: Thursday, March 15, 2012 2:23 PM > To: SPSSX-L@.UGA > Subject: Re: multiple imputation:pooled data > > What is a "pooled dataset"? The way multiple imputation typically works > is that you run your model multiple times, once per imputed data set, and > then work out pooled estimates of the model parameters and standard errors > (using methods described by Donald Rubin). That same basic approach is > used for HLM, I should think. Do you know something that suggests > otherwise? > > HTH. > > > > mariaspinelli wrote >> >> Hi >> >> I'm doing a multiple imputation for missing data. i have to use these >> data for HLM later. I need the pooled dataset in order to use it >> instead that the different single imputations. How can I obtain the >> pooled dataset? >> >> THANK YOU VERY MUCH >> >> >> >> maria >> > > > ----- > -- > Bruce Weaver > bweaver@ > 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/multiple-imputation-pooled-data-tp5568656p5569041.html > Sent from the SPSSX Discussion mailing list archive at Nabble.com. > > ===================== > To manage your subscription to SPSSX-L, send a message to LISTSERV@.UGA > (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 > LISTSERV@.UGA (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 > ----- -- 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/multiple-imputation-pooled-data-tp5568656p5571566.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 |
Re: multiple imputation:pooled data
|
|
In reply to this post by mariaspinelli
So is it your intent to analyze the data in SPSS instead? If your advisor wants to analyze the data in HLM, and has asked you to generate the MDM files, then
you can’t do this in SPSS. Here is some example code for creating the data in SPSS: DATASET DECLARE Test. MULTIPLE IMPUTATION q1 q2 q3 q4
/IMPUTE METHOD=AUTO NIMPUTATIONS=5 MAXPCTMISSING=10 /MISSINGSUMMARIES NONE
/IMPUTATIONSUMMARIES MODELS
/OUTFILE IMPUTATIONS=Test . This will create a new data set that contains 5 imputations, and will be output into the file “test”. Note that this is a very rudimentary version of the Multiple imputation
syntax, and that you have a large number of things to test and resolve before setting this stage up. SAVE OUTFILE='C:\project\Data\test.sav' /COMPRESSED. This leaves you with a stacked file (tall) in which 0 (original) through 5 (5th imputation) are stacked on top of each other (as indicated by a new variable
“imputation_”. In whatever program you use, you would do the analysis “by” the imputation number, and then pool the coefficients at the end, not the data set. In SPSS, you do this through a split file command (Thanks again Bruce for pointing out this possibility
to me).
SORT CASES BY Imputation_. SPLIT FILE LAYERED BY Imputation_. In HLM, I’d have to relook up the syntax to perform this exact task, but I’m sure you can do the same, that should give you what you need. HLM takes the MDM files for
each level (you perform the multiple imputation on each data level separately in this scenario).
Matthew J Poes Research Data Specialist Center for Prevention Research and Development University of Illinois 510 Devonshire Dr. Champaign, IL 61820 Phone: 217-265-4576 email:
[hidden email] From: SPSSX(r) Discussion [mailto:[hidden email]]
On Behalf Of mariaspinelli Thank all. SAS will certainly do it using the by replicate subcommand. I think SPSS has a similar procedure.
If you reply to this email, your message will be added to the discussion below: http://spssx-discussion.1045642.n5.nabble.com/multiple-imputation-pooled-data-tp5568656p5569205.html
To unsubscribe from multiple imputation:pooled data,
click here. View this message in context:
Re: multiple imputation:pooled data |
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |