estimablity assessment
Locked
 
|
On an analysis sequence, I’ve needed to construct planned contrasts using the test subcommand in mixed. Periodically, I’d make mistakes in setting up the L matrix and get the message about estimability. My question is this: how is estimability
assessed/computed? I’ve looked at wiki and stats websites and they’re just not helpful.
Thanks, Gene Maguin |
Administrator
|
Hi Gene. Is this one of the websites you looked at?
http://www-01.ibm.com/support/docview.wss?uid=swg21491942 Can you provide an example (syntax & output) that demonstrates the problem?
--
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/). |
|
|
Hi Bruce,
Now that I look at that page, I'll just bet anything that I have looked at that page before (because it looks familiar) and that somebody, maybe you, maybe Ryan, have pointed out that should I look at it (because I wouldn't have found it otherwise). I don't have an example of where the test command is not working because last week I got all of them to work. But here is a simple one that was giving me trouble last week. What you need to know: Gender (0,1; men=1); powerr1 (1-3; 1=PD, 2=RD, 3=EQ) MIXED relat_stability BY Gender PowerR1 WITH SESchild Relat_Commit AgeRelatStart/FIXED=Gender PowerR1 Gender*PowerR1 SESchild Relat_Commit AgeRelatStart/ PRINT=SOLUTION/RANDOM INTERCEPT | SUBJECT(IDCODE) COVTYPE(ID)/ EMMEANS=TABLES(Gender*PowerR1) WITH (SESchild=mean Relat_Commit=0 AGERELATSTART=20) COMPARE(PowerR1)/ TEST='Women: EQ-PD=0' ALL 0 0 0 1 0 -1 1 0 -1 0 0 0 0 0 0/ TEST='Men: EQ-PD=0' ALL 0 0 0 1 0 -1 0 0 0 1 0 -1 0 0 0/ TEST='Women-Men: EQ-PD=0' ALL 0 0 0 0 0 0 1 0 -1 -1 0 1 0 0 0. I added the lmatrix keyword to the above model statement and it just reprints the Test command coefficients in terms of the parameter table. Not so informative. But, I switched the model to GLM and specified GEF in the print command and out comes the GEF table, all 11 columns. What's interesting is that the second and third test commands are elements of that GEF table but the first is not. I got it that those GEFs are important and they matter but why this set of 11, in the sense of what properties do they satisfy; how those 11 relate to all possible test statements. Don't know. Gene Maguin -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Bruce Weaver Sent: Monday, June 20, 2016 2:20 PM To: [hidden email] Subject: Re: estimablity assessment Hi Gene. Is this one of the websites you looked at? http://www-01.ibm.com/support/docview.wss?uid=swg21491942 Can you provide an example (syntax & output) that demonstrates the problem? Maguin, Eugene wrote > On an analysis sequence, I've needed to construct planned contrasts > using the test subcommand in mixed. Periodically, I'd make mistakes in > setting up the L matrix and get the message about estimability. My > question is > this: how is estimability assessed/computed? I've looked at wiki and > stats websites and they're just not helpful. > Thanks, Gene Maguin > > ===================== > 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/estimablity-assessment-tp5732496p5732498.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 |
|
Administrator
|
Hi Gene. For anything beyond the simplest contrast, I always work things out in Excel. I start with two contrasts that give the fitted values I wish to compare. For these, I fill in the 1s and 0s for first order terms, and program in any product terms as formulae. When I have both vectors, I subtract one from the other.
For your first contrast ('Women: EQ-PD=0'), for example, I got this (view with fixed font): Int G0 G1 P1 P2 P3 G0P1 G0P2 G0P3 G1P1 G1P2 G1P3 SES Relat AgeRelSt C1 1 1 0 0 0 1 0 0 1 0 0 0 0 0 0 C2 1 1 0 1 0 0 1 0 0 0 0 0 0 0 0 Diff 0 0 0 -1 0 1 -1 0 1 0 0 0 0 0 0 C1: Women EQ (Gender=0, PowerR1=3) C2: Women PD (Gender=0, PowerR1=1) Diff: Women: EQ-PD=0 My contrast coefficients are the same as yours, except that the signs are reversed. Mine: 0 0 0 -1 0 1 -1 0 1 0 0 0 0 0 0 Yours: 0 0 0 1 0 -1 1 0 -1 0 0 0 0 0 0 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/). |
|
|
Bruce, thank you. I was thinking in terms of the difference between the equations for predicted values (Women PD - Women EQ) but I wasn't doing it as clearly as you show, which no doubt caused me the trouble I experienced. Gene Maguin
-----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Bruce Weaver Sent: Monday, June 20, 2016 10:13 PM To: [hidden email] Subject: Re: estimablity assessment Hi Gene. For anything beyond the simplest contrast, I always work things out in Excel. I start with two contrasts that give the fitted values I wish to compare. For these, I fill in the 1s and 0s for first order terms, and program in any product terms as formulae. When I have both vectors, I subtract one from the other. For your first contrast ('Women: EQ-PD=0'), for example, I got this (view with fixed font): Int G0 G1 P1 P2 P3 G0P1 G0P2 G0P3 G1P1 G1P2 G1P3 SES Relat AgeRelSt C1 1 1 0 0 0 1 0 0 1 0 0 0 0 0 0 C2 1 1 0 1 0 0 1 0 0 0 0 0 0 0 0 Diff 0 0 0 -1 0 1 -1 0 1 0 0 0 0 0 0 C1: Women EQ (Gender=0, PowerR1=3) C2: Women PD (Gender=0, PowerR1=1) Diff: Women: EQ-PD=0 My contrast coefficients are the same as yours, except that the signs are reversed. Mine: 0 0 0 -1 0 1 -1 0 1 0 0 0 0 0 0 Yours: 0 0 0 1 0 -1 1 0 -1 0 0 0 0 0 0 HTH. Maguin, Eugene wrote > Hi Bruce, > Now that I look at that page, I'll just bet anything that I have > looked at that page before (because it looks familiar) and that > somebody, maybe you, maybe Ryan, have pointed out that should I look > at it (because I wouldn't have found it otherwise). I don't have an > example of where the test command is not working because last week I > got all of them to work. But here is a simple one that was giving me > trouble last week. What you need to know: Gender (0,1; men=1); powerr1 > (1-3; 1=PD, 2=RD, 3=EQ) > > MIXED relat_stability BY Gender PowerR1 WITH SESchild Relat_Commit > AgeRelatStart/FIXED=Gender PowerR1 Gender*PowerR1 SESchild > Relat_Commit AgeRelatStart/ > PRINT=SOLUTION/RANDOM INTERCEPT | SUBJECT(IDCODE) COVTYPE(ID)/ > EMMEANS=TABLES(Gender*PowerR1) WITH (SESchild=mean Relat_Commit=0 > AGERELATSTART=20) COMPARE(PowerR1)/ > TEST='Women: EQ-PD=0' ALL 0 0 0 1 0 -1 1 0 -1 0 0 0 0 0 0/ > TEST='Men: EQ-PD=0' ALL 0 0 0 1 0 -1 0 0 0 1 0 -1 0 0 0/ > TEST='Women-Men: EQ-PD=0' ALL 0 0 0 0 0 0 1 0 -1 -1 0 1 0 0 0. > > I added the lmatrix keyword to the above model statement and it just > reprints the Test command coefficients in terms of the parameter table. > Not so informative. But, I switched the model to GLM and specified GEF > in the print command and out comes the GEF table, all 11 columns. > What's interesting is that the second and third test commands are > elements of that GEF table but the first is not. I got it that those > GEFs are important and they matter but why this set of 11, in the > sense of what properties do they satisfy; how those 11 relate to all > possible test statements. Don't know. > > Gene Maguin > > > > > > -----Original Message----- > From: SPSSX(r) Discussion [mailto: > SPSSX-L@.UGA > ] On Behalf Of Bruce Weaver > Sent: Monday, June 20, 2016 2:20 PM > To: > SPSSX-L@.UGA > Subject: Re: estimablity assessment > > Hi Gene. Is this one of the websites you looked at? > > http://www-01.ibm.com/support/docview.wss?uid=swg21491942 > > Can you provide an example (syntax & output) that demonstrates the > problem? > > > > > Maguin, Eugene wrote >> On an analysis sequence, I've needed to construct planned contrasts >> using the test subcommand in mixed. Periodically, I'd make mistakes >> in setting up the L matrix and get the message about estimability. My >> question is >> this: how is estimability assessed/computed? I've looked at wiki and >> stats websites and they're just not helpful. >> Thanks, Gene Maguin >> >> ===================== >> 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 > 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/estimablity-assessment-t > p5732496p5732498.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/estimablity-assessment-tp5732496p5732500.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 |