SPSS MATRIX - LIMITATION
Locked
 
|
Dear SPSS-L (and especially those who work for IBM-SPSS), When responding to posts related to all statistical models which fall under the umbrella of generalized linear mixed models, when time permits, I try to provide simulation code to: (a) help the OP fill in the missing pieces that I simply do not have time to explain and (b) validate the claimed parameterization of the model underlying the statistical procedure code that is offered (e.g., MIXED).
Generating data that follow a multivariate normal distribution with a specified covariance structure (which is obviously common with my simulation experiments) would be very easy if something analogous to rv.normal were offered within MATRIX.
1. Why is this capability not currently offered? 2. Are there plans in upcoming releases to include this capability? If not, I would ask those who work for IBM - SPSS to note this request from a loyal customer who has been using this software in industry and academia on a regular basis for many years.
Note that the MATRIX module (IML) in SAS, the other software I use routinely, offers the option of generating data with a specified covariance matrix from a multivariate random normal distribution: Thanks, Ryan
|
Administrator
|
Don't hold your breath, Ryan. Why not? Exhibit 1:
http://spssx-discussion.1045642.n5.nabble.com/Inclusion-of-IDF-functions-in-matrix-language-td4680040.html 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 Ryan
The Make New Dataset with Cases custom
dialog box available from the SPSS Community site generates random data
from 23 different distributions and any number of variables and cases,
including normal, and you can rotate the data into any correlation structure
you want: equal correlations, Toeplitz, factor analytic, arbitrary where
you specify all the correlations, and random. It requires the Python
Essentials. It is available in the Utilities Collection.
The dialog generates an input program and, for correlations, a MATRIX command. Jon Peck (no "h") aka Kim Senior Software Engineer, IBM [hidden email] phone: 720-342-5621 From: R B <[hidden email]> To: [hidden email], Date: 03/24/2013 11:45 AM Subject: [SPSSX-L] SPSS MATRIX - LIMITATION Sent by: "SPSSX(r) Discussion" <[hidden email]> Dear SPSS-L (and especially those who work for IBM-SPSS), When responding to posts related to all statistical models which fall under the umbrella of generalized linear mixed models, when time permits, I try to provide simulation code to: (a) help the OP fill in the missing pieces that I simply do not have time to explain and (b) validate the claimed parameterization of the model underlying the statistical procedure code that is offered (e.g., MIXED). Generating data that follow a multivariate normal distribution with a specified covariance structure (which is obviously common with my simulation experiments) would be very easy if something analogous to rv.normal were offered within MATRIX. 1. Why is this capability not currently offered? 2. Are there plans in upcoming releases to include this capability? If not, I would ask those who work for IBM - SPSS to note this request from a loyal customer who has been using this software in industry and academia on a regular basis for many years. Note that the MATRIX module (IML) in SAS, the other software I use routinely, offers the option of generating data with a specified covariance matrix from a multivariate random normal distribution: http://support.sas.com/documentation/cdl/en/imlug/64248/HTML/default/viewer.htm#imlug_modlib_sect020.htm Thanks, Ryan |
|
|
I meant to add that if you have version
21, you can use the new simulation capabilities - a Base feature
- to generate variables from a large number of distributions, and you can
specify a correlation matrix or, alternatively, estimate one from the active
dataset. You can simulate a saved pmml mode or just type in an equation.
The system will generate the input data according to your specifications
and can save that for future use.
Jon Peck (no "h") aka Kim Senior Software Engineer, IBM [hidden email] phone: 720-342-5621 From: Jon K Peck/Chicago/IBM@IBMUS To: [hidden email], Date: 03/24/2013 01:50 PM Subject: Re: [SPSSX-L] SPSS MATRIX - LIMITATION Sent by: "SPSSX(r) Discussion" <[hidden email]> The Make New Dataset with Cases custom dialog box available from the SPSS Community site generates random data from 23 different distributions and any number of variables and cases, including normal, and you can rotate the data into any correlation structure you want: equal correlations, Toeplitz, factor analytic, arbitrary where you specify all the correlations, and random. It requires the Python Essentials. It is available in the Utilities Collection. The dialog generates an input program and, for correlations, a MATRIX command. Jon Peck (no "h") aka Kim Senior Software Engineer, IBM [hidden email] phone: 720-342-5621 From: R B <[hidden email]> To: [hidden email], Date: 03/24/2013 11:45 AM Subject: [SPSSX-L] SPSS MATRIX - LIMITATION Sent by: "SPSSX(r) Discussion" <[hidden email]> Dear SPSS-L (and especially those who work for IBM-SPSS), When responding to posts related to all statistical models which fall under the umbrella of generalized linear mixed models, when time permits, I try to provide simulation code to: (a) help the OP fill in the missing pieces that I simply do not have time to explain and (b) validate the claimed parameterization of the model underlying the statistical procedure code that is offered (e.g., MIXED). Generating data that follow a multivariate normal distribution with a specified covariance structure (which is obviously common with my simulation experiments) would be very easy if something analogous to rv.normal were offered within MATRIX. 1. Why is this capability not currently offered? 2. Are there plans in upcoming releases to include this capability? If not, I would ask those who work for IBM - SPSS to note this request from a loyal customer who has been using this software in industry and academia on a regular basis for many years. Note that the MATRIX module (IML) in SAS, the other software I use routinely, offers the option of generating data with a specified covariance matrix from a multivariate random normal distribution: http://support.sas.com/documentation/cdl/en/imlug/64248/HTML/default/viewer.htm#imlug_modlib_sect020.htm Thanks, Ryan |
Administrator
|
In reply to this post by Ryan
Thanks for posting this Ryan,
See my Dec 10 posting (and the ensuing thread/bloodbath for a bit of background indicating you are not alone). http://spssx-discussion.1045642.n5.nabble.com/The-agony-and-the-ecstacy-Development-of-complex-MATRIX-programs-td5716810.html#a5716830 Semi Ugly Workaround : 1. Generate the vectors within the TAFKASPSS transformation language (TL). 2. Handoff the active file to MATRIX to compute anything you can't do with TL . 3. Handoff the matrix computed elements back to YAFKASPSS to evaluate whatever else you require. This is the approach I am taking presently to get the NCDF.T required for some power calculations. Using MATRIX to compute the df and ncp for some pretty complicated designs. David ---
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?" |
|
In reply to this post by Ryan
Does the SIMULATION
procedure have an option to
generate a matrix with a given covariance
structure?
Check the archives, IIRC, there have been posts that use the CHOLESKY to do this. Art Kendall Social Research ConsultantsOn 3/24/2013 1:45 PM, SUBSCRIBE SPSSX-L Anonymous [via SPSSX Discussion] wrote:
Art Kendall
Social Research Consultants |
|
Administrator
|
"Check the archives, IIRC, there have been posts that use the CHOLESKY to do this. "
http://spssx-discussion.1045642.n5.nabble.com/template/NamlServlet.jtp?macro=search_page&node=1068821&query=CHOL&n=1068821 I posted most of these (in fact posted code this AM which makes the 'limitation' somewhat a non issue since one can toggle back and forth between MATRIX and SPSS base - I STILL WOULDN'T MIND seeing the neglected red-headed b@$^@&|) get a little attention). Meanwhile I have determined that that code doesn't do exactly what Ryan's neg var comp simulation explores and we are in off list communication re how best to resolve this issue. So, count on an update to that.
Please reply to the list and not to my personal email.
Those desiring my consulting or training services please feel free to email me. --- "Nolite dare sanctum canibus neque mittatis margaritas vestras ante porcos ne forte conculcent eas pedibus suis." Cum es damnatorum possederunt porcos iens ut salire off sanguinum cliff in abyssum?" |
|
|
For some reason, it looks as though I haven't received Art K's emails
in a while. Not sure why. Anyway, if Art happens to see this email, I'd like to let him know that I'm investigating the potential capabilities of this new procedure with respect to various var-cov residual matrices. I think Jon P. said the SIMULATION procedure is capable of generating data from various variance-covariance matrices. Ryan On Thu, Mar 28, 2013 at 12:44 PM, David Marso <[hidden email]> wrote: > "Check the archives, IIRC, there have been posts that use the CHOLESKY to do > this. " > > http://spssx-discussion.1045642.n5.nabble.com/template/NamlServlet.jtp?macro=search_page&node=1068821&query=CHOL&n=1068821 > > I posted most of these (in fact posted code this AM which makes the > 'limitation' somewhat a non issue since one can toggle back and forth > between MATRIX and SPSS base - I STILL WOULDN'T MIND seeing the neglected > red-headed b@$^@&|) get a little attention). > Meanwhile I have determined that that code doesn't do exactly what Ryan's > neg var comp simulation explores and we are in off list communication re how > best to resolve this issue. > So, count on an update to that. > > > > Art Kendall wrote >> Does the SIMULATION >> procedure have an option to >> generate a matrix with a given covariance >> structure? >> >> Check the archives, IIRC, there have >> been posts that use the CHOLESKY to do this. >> Art Kendall >> Social Research Consultants >> On 3/24/2013 1:45 PM, SUBSCRIBE SPSSX-L Anonymous [via SPSSX >> Discussion] wrote: >> >> >> Dear SPSS-L (and especially those who work for IBM-SPSS), >>  >> When responding to posts related to all statistical models >> which fall under the umbrella of generalized linear mixed >> models, when time permits, I try to provide simulation code to: >> (a) help the OP fill in the missing pieces that I simply do not >> have time to explain and (b) validate the claimed >> parameterization of the model underlying the statistical >> procedure code that is offered (e.g., MIXED). >>  >> Generating data that follow a multivariate normal >> distribution with a specified covariance structure (which is >> obviously common with my simulation experiments) would be very >> easy if something analogous to rv.normal were offered within >> MATRIX. >>  >> 1. Why is this capability not currently offered? >>  >> 2. Are there plans in upcoming releases to include this >> capability? >>  >> If not, I would ask those who work for IBM - SPSS to note >> this request from a loyal customer who has been using >> this software in industry and academia on a regular basis for >> many years.  >>  >> Note that the MATRIX module (IML) in SAS, the other software >> I use routinely, offers the option of generating data with a >> specified covariance matrix from a multivariate random normal >> distribution: >> >>  >> >> http://support.sas.com/documentation/cdl/en/imlug/64248/HTML/default/viewer.htm#imlug_modlib_sect020.htm >>  >> Thanks, >>  >> Ryan >> >> >> >> >> If you reply to this email, your >> message will be added to the discussion below: >> >> http://spssx-discussion.1045642.n5.nabble.com/SPSS-MATRIX-LIMITATION-tp5719047.html >> >> >> To start a new topic under SPSSX Discussion, email >> > >> ml-node+s1045642n1068821h68@.nabble > >> >> To unsubscribe from SPSSX Discussion, click >> here . >> NAML > > > > > > ----- > 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?" > -- > View this message in context: http://spssx-discussion.1045642.n5.nabble.com/SPSS-MATRIX-LIMITATION-tp5719047p5719143.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 |