My question concerns two separate tables of "Multivariate" tests produced by the GLM command. As an example, I will use a design with two groups (e.g., males/females). Each individual was measured at each of three time points (time1/time2/time3) on each of four dependent variables (dv1/dv2/dv3/dv4). (At the end of this question is a syntax file that will generate sample data and run the analysis, allowing me to ask the question with reference to specific values.) The first "Multivariate Tests^a" output table appears early in the output, and it includes one section for "Between Subjects" effects (i.e., intercept & group, in this design) and another section for "Within Subjects" effects (i.e., Time and Time*group). The second "Multivariate^a,b" table appears later in the output (after Mauchley's test) under the major heading "Tests of WithinSubjects Effects". Each of these two tables displays the results of multivariate tests (eg Pillai's Trace, Wilks' lambda) for the two WithinSubjects sources (i.e., Time and Time*group), but the two tables give different values. My question is simply, "what is the difference between these two tables?" For example, what are the models and null hypotheses in each case? Why are the error df's different? What would it mean if the first table's test of Time was significant and the second table's was not, or vice versa? Any insight would be greatly appreciated. Thanks for your time.  Syntax file follows: Just copy, paste, and run all to see a specific data set and analysis illustrating my question. * Here is a syntax file to generate and analyze an example data set, * so that the question can be posed in terms of specific values. * For this example, there are 2 groups of randomlyselected individuals (cases). * Each case is measured at 3 time points on 4 different DVs. * Set the RNG seed so that the results are reproduced exactly. set RNG=MT MTINDEX=121314.15. new file. input program. compute #SampleSize = 50. loop #i = 1 to #SampleSize. compute group = 1. compute dv1time1 = rv.normal(0,1). compute dv1time2 = rv.normal(0,1). compute dv1time3 = rv.normal(0,1). compute dv2time1 = rv.normal(0,1). compute dv2time2 = rv.normal(0,1). compute dv2time3 = rv.normal(0,1). compute dv3time1 = rv.normal(0,1). compute dv3time2 = rv.normal(0,1). compute dv3time3 = rv.normal(0,1). compute dv4time1 = rv.normal(0,1). compute dv4time2 = rv.normal(0,1). compute dv4time3 = rv.normal(0,1). end case. end loop. loop #i = 1 to #SampleSize. compute group = 2. compute dv1time1 = rv.normal(0,1). compute dv1time2 = rv.normal(0,1). compute dv1time3 = rv.normal(0,1). compute dv2time1 = rv.normal(0,1). compute dv2time2 = rv.normal(0,1). compute dv2time3 = rv.normal(0,1). compute dv3time1 = rv.normal(0,1). compute dv3time2 = rv.normal(0,1). compute dv3time3 = rv.normal(0,1). compute dv4time1 = rv.normal(0,1). compute dv4time2 = rv.normal(0,1). compute dv4time3 = rv.normal(0,1). end case. end loop. end file. end input program. execute. GLM dv1time1 dv1time2 dv1time3 dv2time1 dv2time2 dv2time3 dv3time1 dv3time2 dv3time3 dv4time1 dv4time2 dv4time3 BY group /WSFACTOR = Time 3 Polynomial /MEASURE = DV1 DV2 DV3 DV4 /METHOD = SSTYPE(3) /EMMEANS = TABLES(group) /EMMEANS = TABLES(Time) /EMMEANS = TABLES(group*Time) /PRINT = DESCRIPTIVE HOMOGENEITY /CRITERIA = ALPHA(.05) /WSDESIGN = Time /DESIGN = group . * Here is a more specific version of my question. * * An earlier "Multivariate Tests^a" output table includes the line: * Within Subjects Time Pillai's Trace 0.145 1.935^b 8.000 91.000 0.064 * * Two tables farther along is a "Multivariate^a,b" table that includes the line: * Within Subjects Time Pillai's Trace 0.068 1.712 8.000 388.000 0.094 * * These two tables include the same sources (i.e., Time and Time*group), * but they give different values and different error df's. * * What is the difference between these two tables? * (My question pertains to both of the sources, Time and Time*group). 
The first table represents Multivariate approach results to
RMANOVA.
The second table (Tests of withinsubject effects) represents Univariate approach results to RMANOVA. SPSS Help for RM GLM clearly outlines the distinction. See also e.g. here http://stats.stackexchange.com/a/15212/3277 with some further links. 05.09.2016 5:12, Jeff Miller пишет:
My question concerns two separate tables of "Multivariate" tests produced by the GLM command. As an example, I will use a design with two groups (e.g., males/females). Each individual was measured at each of three time points (time1/time2/time3) on each of four dependent variables (dv1/dv2/dv3/dv4). (At the end of this question is a syntax file that will generate sample data and run the analysis, allowing me to ask the question with reference to specific values.) The first "Multivariate Tests^a" output table appears early in the output, and it includes one section for "Between Subjects" effects (i.e., intercept & group, in this design) and another section for "Within Subjects" effects (i.e., Time and Time*group). The second "Multivariate^a,b" table appears later in the output (after Mauchley's test) under the major heading "Tests of WithinSubjects Effects". Each of these two tables displays the results of multivariate tests (eg Pillai's Trace, Wilks' lambda) for the two WithinSubjects sources (i.e., Time and Time*group), but the two tables give different values. My question is simply, "what is the difference between these two tables?" For example, what are the models and null hypotheses in each case? Why are the error df's different? What would it mean if the first table's test of Time was significant and the second table's was not, or vice versa? Any insight would be greatly appreciated. Thanks for your time.  Syntax file follows: Just copy, paste, and run all to see a specific data set and analysis illustrating my question. * Here is a syntax file to generate and analyze an example data set, * so that the question can be posed in terms of specific values. * For this example, there are 2 groups of randomlyselected individuals (cases). * Each case is measured at 3 time points on 4 different DVs. * Set the RNG seed so that the results are reproduced exactly. set RNG=MT MTINDEX=121314.15. new file. input program. compute #SampleSize = 50. loop #i = 1 to #SampleSize. compute group = 1. compute dv1time1 = rv.normal(0,1). compute dv1time2 = rv.normal(0,1). compute dv1time3 = rv.normal(0,1). compute dv2time1 = rv.normal(0,1). compute dv2time2 = rv.normal(0,1). compute dv2time3 = rv.normal(0,1). compute dv3time1 = rv.normal(0,1). compute dv3time2 = rv.normal(0,1). compute dv3time3 = rv.normal(0,1). compute dv4time1 = rv.normal(0,1). compute dv4time2 = rv.normal(0,1). compute dv4time3 = rv.normal(0,1). end case. end loop. loop #i = 1 to #SampleSize. compute group = 2. compute dv1time1 = rv.normal(0,1). compute dv1time2 = rv.normal(0,1). compute dv1time3 = rv.normal(0,1). compute dv2time1 = rv.normal(0,1). compute dv2time2 = rv.normal(0,1). compute dv2time3 = rv.normal(0,1). compute dv3time1 = rv.normal(0,1). compute dv3time2 = rv.normal(0,1). compute dv3time3 = rv.normal(0,1). compute dv4time1 = rv.normal(0,1). compute dv4time2 = rv.normal(0,1). compute dv4time3 = rv.normal(0,1). end case. end loop. end file. end input program. execute. GLM dv1time1 dv1time2 dv1time3 dv2time1 dv2time2 dv2time3 dv3time1 dv3time2 dv3time3 dv4time1 dv4time2 dv4time3 BY group /WSFACTOR = Time 3 Polynomial /MEASURE = DV1 DV2 DV3 DV4 /METHOD = SSTYPE(3) /EMMEANS = TABLES(group) /EMMEANS = TABLES(Time) /EMMEANS = TABLES(group*Time) /PRINT = DESCRIPTIVE HOMOGENEITY /CRITERIA = ALPHA(.05) /WSDESIGN = Time /DESIGN = group . * Here is a more specific version of my question. * * An earlier "Multivariate Tests^a" output table includes the line: * Within Subjects Time Pillai's Trace 0.145 1.935^b 8.000 91.000 0.064 * * Two tables farther along is a "Multivariate^a,b" table that includes the line: * Within Subjects Time Pillai's Trace 0.068 1.712 8.000 388.000 0.094 * * These two tables include the same sources (i.e., Time and Time*group), * but they give different values and different error df's. * * What is the difference between these two tables? * (My question pertains to both of the sources, Time and Time*group).  View this message in context: http://spssxdiscussion.1045642.n5.nabble.com/QuestionaboutSPSSMANOVAoutputtp5733053.html Sent from the SPSSX Discussion mailing list archive at Nabble.com. ===================== To manage your subscription to SPSSXL, send a message to [hidden email] (not to SPSSXL), with no body text except the command. To leave the list, send the command SIGNOFF SPSSXL For a list of commands to manage subscriptions, send the command INFO REFCARD
===================== To manage your subscription to SPSSXL, send a message to [hidden email] (not to SPSSXL), with no body text except the command. To leave the list, send the command SIGNOFF SPSSXL For a list of commands to manage subscriptions, send the command INFO REFCARD 
Administrator

That doesn't sound right to me, Kirill. As the attached Excel file shows, both tables Jeff refers to are showing multivariate test results for the Time and Time*Group terms (with no separation by DV1 through DV4).
Members who do not use Nabble can download the attached Excel file here: http://spssxdiscussion.1045642.n5.nabble.com/QuestionaboutSPSSMANOVAoutputtd5733053.html Miller_example.xls

Bruce Weaver bweaver@lakeheadu.ca 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 email, please use the address shown above. 
Okay, believe it or not, this question has been asked previously
(on the Usenet spss group and the answer is on the IBM SPSS website). The key thing to notice is that the second table has the footnote: b. Tests based on averaged variables This is makes the second Multivariate results different from the first table which uses the original variables. For more on this point see: https://groups.google.com/forum/#!topic/comp.softsys.stat.spss/UeqpN6ISxy8 and http://www01.ibm.com/support/docview.wss?uid=swg21476707 If the withinsubject factor has only two levels, both multivariate results would be the same. The following quote is from the IBM website in describing why the two multivariate results differ when WS levels > 2 (and more than one dependent variable is used):  The first table entitled Multivariate Tests carries the true doublymultivariate test statistics. The averaged results appear in the table entitled Multivariate that is listed under the heading Tests of WithinSubjects Effects. Averaged results are like those you would have obtained had you set up a multivariate model with a data set that is configured for the univariate approach to repeated measures (e.g., a splitplot approach). The IBM SPSS webiste provides three GLM syntax programs that shows how one would do the doublymultivariate analysis (first syntax) and how to do the univariate or splitplot analysis (3rd syntax; NOTE: "Subject" is included as a factor for the example used  for the OP's situation, he doesn't really have a doubly multivariate situation but a "straight" univariate withnsubject design requiring a repeated measures ANOVA). The splitplot analysis assumes a form of multivariate sphericity that SPSS does not text for, hence, one wouldn't know if the analysis appropriate or the degree of the violation. The doubly multivariate analysis does not make this assumption, hence one would focus on the first set of Multivariate results if that is of concern/interest. Now, I assume that that the OP just wants an oldfashioned Gddamned repeated measues ANOVA and is not really interested in the Multivariate or MANOVA (not to be confused with the SPSS procedure MANOVA) results. If printing the Multivariate results were left up to the programmer instead of being automatically presented, the confusion that invariably shows up (NOTE: the Usenet group post is from 2009 and it references older posts asking the question "Why oh why do the Multivariate results tables disagree?!?"). It would be helpful if (a) the distinction provided on IBM SPSS webpage was clearly provided in the SPSS manual (I looked but didn't find it; I'm open to correction on this point) and/or (b) provided a brief explanation in the output after the second Multivariate table. Or maybe they could build in a command or keyword "repeated" into GLM that it should *only* be treated as a repeated measures design instead of some kind of multivariate analysis. This would save having to delete the multivariate results especially if one does a oneway repeated measures ANOVA. But what do I know, right? Mike Palij New York University [hidden email]  Original Message  From: "Bruce Weaver" <[hidden email]> To: <[hidden email]> Sent: Tuesday, September 06, 2016 9:12 AM Subject: Re: Question about SPSS MANOVA output That doesn't sound right to me, Kirill. As the attached Excel file shows, *both* tables Jeff refers to are showing multivariate test results for the Time and Time*Group terms (with no separation by DV1 through DV4). Members who do not use Nabble can download the attached Excel file here: http://spssxdiscussion.1045642.n5.nabble.com/QuestionaboutSPSSMANOVAoutputtd5733053.html Miller_example.xls <http://spssxdiscussion.1045642.n5.nabble.com/file/n5733058/Miller_example.xls> Kirill Orlov wrote > The first table represents Multivariate approach results to RMANOVA. > The second table (Tests of withinsubject effects) represents > Univariate > approach results to RMANOVA. > SPSS Help for RM GLM clearly outlines the distinction. > See also e.g. here http://stats.stackexchange.com/a/15212/3277 with > some > further links. > > 05.09.2016 5:12, Jeff Miller пишет: >> My question concerns two separate tables of "Multivariate" tests >> produced >> by >> the GLM command. As an example, I will use a design with two groups >> (e.g., >> males/females). Each individual was measured at each of three time >> points >> (time1/time2/time3) on each of four dependent variables >> (dv1/dv2/dv3/dv4). >> >> (At the end of this question is a syntax file that will generate >> sample >> data >> and run the analysis, allowing me to ask the question with reference >> to >> specific values.) >> >> The first "Multivariate Tests^a" output table appears early in the >> output, >> and it includes one section for "Between Subjects" effects (i.e., >> intercept >> & group, in this design) and another section for "Within Subjects" >> effects >> (i.e., Time and Time*group). >> >> The second "Multivariate^a,b" table appears later in the output >> (after >> Mauchley's test) under the major heading "Tests of WithinSubjects >> Effects". >> >> Each of these two tables displays the results of multivariate tests >> (eg >> Pillai's Trace, Wilks' lambda) for the two WithinSubjects sources >> (i.e., >> Time and Time*group), but the two tables give different values. >> >> My question is simply, "what is the difference between these two >> tables?" >> For example, what are the models and null hypotheses in each case? >> Why >> are >> the error df's different? What would it mean if the first table's >> test >> of >> Time was significant and the second table's was not, or vice versa? >> >> Any insight would be greatly appreciated. Thanks for your time. >> >>  Syntax file follows: Just copy, paste, and run all to see a >> specific >> data set and analysis illustrating my question. >> >> * Here is a syntax file to generate and analyze an example data set, >> * so that the question can be posed in terms of specific values. >> >> * For this example, there are 2 groups of randomlyselected >> individuals >> (cases). >> * Each case is measured at 3 time points on 4 different DVs. >> >> * Set the RNG seed so that the results are reproduced exactly. >> set RNG=MT MTINDEX=121314.15. >> >> new file. >> input program. >> >> compute #SampleSize = 50. >> >> loop #i = 1 to #SampleSize. >> compute group = 1. >> compute dv1time1 = rv.normal(0,1). >> compute dv1time2 = rv.normal(0,1). >> compute dv1time3 = rv.normal(0,1). >> compute dv2time1 = rv.normal(0,1). >> compute dv2time2 = rv.normal(0,1). >> compute dv2time3 = rv.normal(0,1). >> compute dv3time1 = rv.normal(0,1). >> compute dv3time2 = rv.normal(0,1). >> compute dv3time3 = rv.normal(0,1). >> compute dv4time1 = rv.normal(0,1). >> compute dv4time2 = rv.normal(0,1). >> compute dv4time3 = rv.normal(0,1). >> end case. >> end loop. >> loop #i = 1 to #SampleSize. >> compute group = 2. >> compute dv1time1 = rv.normal(0,1). >> compute dv1time2 = rv.normal(0,1). >> compute dv1time3 = rv.normal(0,1). >> compute dv2time1 = rv.normal(0,1). >> compute dv2time2 = rv.normal(0,1). >> compute dv2time3 = rv.normal(0,1). >> compute dv3time1 = rv.normal(0,1). >> compute dv3time2 = rv.normal(0,1). >> compute dv3time3 = rv.normal(0,1). >> compute dv4time1 = rv.normal(0,1). >> compute dv4time2 = rv.normal(0,1). >> compute dv4time3 = rv.normal(0,1). >> end case. >> end loop. >> >> end file. >> end input program. >> execute. >> >> GLM >> dv1time1 dv1time2 dv1time3 dv2time1 dv2time2 dv2time3 >> dv3time1 dv3time2 dv3time3 dv4time1 dv4time2 dv4time3 BY group >> /WSFACTOR = Time 3 Polynomial >> /MEASURE = DV1 DV2 DV3 DV4 >> /METHOD = SSTYPE(3) >> /EMMEANS = TABLES(group) >> /EMMEANS = TABLES(Time) >> /EMMEANS = TABLES(group*Time) >> /PRINT = DESCRIPTIVE HOMOGENEITY >> /CRITERIA = ALPHA(.05) >> /WSDESIGN = Time >> /DESIGN = group . >> >> * Here is a more specific version of my question. >> * >> * An earlier "Multivariate Tests^a" output table includes the line: >> * Within Subjects Time Pillai's Trace 0.145 >> 1.935^b 8.000 91.000 0.064 >> * >> * Two tables farther along is a "Multivariate^a,b" table that >> includes >> the >> line: >> * Within Subjects Time Pillai's Trace 0.068 >> 1.712 8.000 388.000 0.094 >> * >> * These two tables include the same sources (i.e., Time and >> Time*group), >> * but they give different values and different error df's. >> * >> * What is the difference between these two tables? >> * (My question pertains to both of the sources, Time and Time*group). >> >> >> >> >> >>  >> View this message in context: >> http://spssxdiscussion.1045642.n5.nabble.com/QuestionaboutSPSSMANOVAoutputtp5733053.html >> Sent from the SPSSX Discussion mailing list archive at Nabble.com. >> >> ===================== >> To manage your subscription to SPSSXL, send a message to >> > LISTSERV@.UGA > (not to SPSSXL), with no body text except the >> command. To leave the list, send the command >> SIGNOFF SPSSXL >> For a list of commands to manage subscriptions, send the command >> INFO REFCARD >> > > > > >  > Это сообщение проверено на вирусы антивирусом Avast. > https://www.avast.com/antivirus > > ===================== > To manage your subscription to SPSSXL, send a message to > LISTSERV@.UGA > (not to SPSSXL), with no body text except the > command. To leave the list, send the command > SIGNOFF SPSSXL > 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 email, please use the address shown above.  View this message in context: http://spssxdiscussion.1045642.n5.nabble.com/QuestionaboutSPSSMANOVAoutputtp5733053p5733058.html Sent from the SPSSX Discussion mailing list archive at Nabble.com. ===================== To manage your subscription to SPSSXL, send a message to [hidden email] (not to SPSSXL), with no body text except the command. To leave the list, send the command SIGNOFF SPSSXL For a list of commands to manage subscriptions, send the command INFO REFCARD ===================== To manage your subscription to SPSSXL, send a message to [hidden email] (not to SPSSXL), with no body text except the command. To leave the list, send the command SIGNOFF SPSSXL For a list of commands to manage subscriptions, send the command INFO REFCARD 
In reply to this post by Jeff Miller
1st output table is the one to use
it makes full use of the correlation matrix between repeated measures, i.e different correlation between t1, t2 and t2,t3 2nd table ASSUMES all pairwise correlation same and estimates a mean correlation and then makes various corrections for correlation differences it was rocket science when SPSS launched some decades ago when the truly multivariate method was computationally expensive why it is still there I have no idea, but spss does not like to drop anything the Maunchley test is testing whether all the pairwise correlations ARE the same. In my view it is better to use full correlation matrix even if Maunchley is ns [power is low anyway] best Diana On 5 Sep 2016, at 03:12, Jeff Miller <[hidden email]> wrote: ________________________________________ Professor Diana Kornbrot Work University of Hertfordshire College Lane, Hatfield, Hertfordshire AL10 9AB, UK +44 (0) 170 728 4626 [hidden email] http://dianakornbrot.wordpress.com/ http://go.herts.ac.uk/Diana_Kornbrot skype: kornbrotme Home 19 Elmhurst Avenue London N2 0LT, UK +44 (0) 208 444 2081 
After doing some more searching of the literature and the web,
let me say that Diana is correct, that is, if one is going to use a multivariate analysis instead of traditional repeated measures ANOVA, the approach SPSS calls "doublymultivariate results" should be used. But a few more points need to be made: (1) The question about the two sets of multivariate results was original asked on the SPSS Usenet group back in 1998 and David Nichols (Hi David!) gave a response. His post can be read here: https://groups.google.com/forum/#!topic/comp.softsys.stat.spss/857xHuoXDYw And his response is reproduced here: I already answered these questions individually, but in case others are interested, the tests mentioned are indeed the same as those labeled averaged multivariate tests in MANOVA. NOTE to those still using the SPSS MANOVA procedure: check to see whether only the average multivariate results are still provided. They ignore the parts of the SSCP matrices that involve covariation among transformed variables within the same measure, but incorporate covariation among transformed variables in different measures to produce multivariate tests. I 've seen them referred to as Scheffe's multivariate mixed model tests in an article by Robert Boik (I don't have the full reference handy). Unfortunately, SPSS does not at the current time have a test for the condition of multisample sphericity that is required for them to be valid. To get corrected averaged univariate tests printed in Release 7.5, double click on the table with the averaged tests, then choose Pivot>Move Layers to Rows, or activate the pivoting trays and move the Epsilon corrections icons from the Layers area to the Rows area. In Release 8.0, the table is presented with these tests displayed. (2) I don't know which Boik article David is referring to but one that explicitly refers to the second method is covered in the following reference: Boik, R. J. (1991). Scheffés mixed model for multivariate repeated measures: a relative efficiency evaluation. Communications in StatisticsTheory and Methods, 20(4), 12331255. At the end of the article's abstract Boik writes: Based on relative efficiency evaluations, the adjusted MMM [NOTE: MMM = Scheffe's multivatiate mixed model] analyses described in this paper can be recommended only when sample sizes are very small or there is reason to believe that multivariate sphericity is nearly satisfied. Neither the €adjusted analysis suggested in the SPSS MANOVA output (release 4.0) nor the adjusted analysis suggested by Boik (1988) can be recommended at all. NOTE: Yes, Boik refers to good old version 4.0 of SPSS. I Don't think SPSS has changed MANOVA or the multivariate analysis much since back then (except GLM came after version 4.0). (3) An old reference that argues strongly for the Multivariate approach to repeated measures analysis is the following reference: O'Brien, R. G., & Kaiser, M. K. (1985). MANOVA method for analyzing repeated measures designs: an extensive primer. Psychological bulletin, 97(2), 316. The authors use the term MANOVA to refer to both the statistical procedure and the SPSSX (what it was called back then) procedure. (4) Given the issues above, one wonders why SPSS has not provided the multivariate sphericity test that would show whether the second set of multivariate (Scheffe) results are valid. Without it, one has to try to work it out on one's own or see if some other package provides it (or do the analysis in a SEM package). And does the MANOVA procedure (still available via syntax) still only give the Scheffe results? If so, warnings about whether to use it or not should be more widely communicated. If anyone has more uptodate knowledge about this situation, either in SPSS or in other programs (e.g., SAS, Stata, etc.), I think it would be of benefit to the user community if it were provided here. Mike Palij New York University [hidden email]  Original Message  From: Kornbrot, Diana To: [hidden email] Sent: Tuesday, September 06, 2016 11:27 AM Subject: Re: Question about SPSS MANOVA output 1st output table is the one to use it makes full use of the correlation matrix between repeated measures, i.e different correlation between t1, t2 and t2,t3 2nd table ASSUMES all pairwise correlation same and estimates a mean correlation and then makes various corrections for correlation differences it was rocket science when SPSS launched some decades ago when the truly multivariate method was computationally expensive why it is still there I have no idea, but spss does not like to drop anything the Maunchley test is testing whether all the pairwise correlations ARE the same. In my view it is better to use full correlation matrix even if Maunchley is ns [power is low anyway] best Diana On 5 Sep 2016, at 03:12, Jeff Miller <[hidden email]> wrote: My question concerns two separate tables of "Multivariate" tests produced by the GLM command. As an example, I will use a design with two groups (e.g., males/females). Each individual was measured at each of three time points (time1/time2/time3) on each of four dependent variables (dv1/dv2/dv3/dv4). (At the end of this question is a syntax file that will generate sample data and run the analysis, allowing me to ask the question with reference to specific values.) The first "Multivariate Tests^a" output table appears early in the output, and it includes one section for "Between Subjects" effects (i.e., intercept & group, in this design) and another section for "Within Subjects" effects (i.e., Time and Time*group). The second "Multivariate^a,b" table appears later in the output (after Mauchley's test) under the major heading "Tests of WithinSubjects Effects". Each of these two tables displays the results of multivariate tests (eg Pillai's Trace, Wilks' lambda) for the two WithinSubjects sources (i.e., Time and Time*group), but the two tables give different values. My question is simply, "what is the difference between these two tables?" For example, what are the models and null hypotheses in each case? Why are the error df's different? What would it mean if the first table's test of Time was significant and the second table's was not, or vice versa? Any insight would be greatly appreciated. Thanks for your time.  Syntax file follows: Just copy, paste, and run all to see a specific data set and analysis illustrating my question. * Here is a syntax file to generate and analyze an example data set, * so that the question can be posed in terms of specific values. * For this example, there are 2 groups of randomlyselected individuals (cases). * Each case is measured at 3 time points on 4 different DVs. * Set the RNG seed so that the results are reproduced exactly. set RNG=MT MTINDEX=121314.15. new file. input program. compute #SampleSize = 50. loop #i = 1 to #SampleSize. compute group = 1. compute dv1time1 = rv.normal(0,1). compute dv1time2 = rv.normal(0,1). compute dv1time3 = rv.normal(0,1). compute dv2time1 = rv.normal(0,1). compute dv2time2 = rv.normal(0,1). compute dv2time3 = rv.normal(0,1). compute dv3time1 = rv.normal(0,1). compute dv3time2 = rv.normal(0,1). compute dv3time3 = rv.normal(0,1). compute dv4time1 = rv.normal(0,1). compute dv4time2 = rv.normal(0,1). compute dv4time3 = rv.normal(0,1). end case. end loop. loop #i = 1 to #SampleSize. compute group = 2. compute dv1time1 = rv.normal(0,1). compute dv1time2 = rv.normal(0,1). compute dv1time3 = rv.normal(0,1). compute dv2time1 = rv.normal(0,1). compute dv2time2 = rv.normal(0,1). compute dv2time3 = rv.normal(0,1). compute dv3time1 = rv.normal(0,1). compute dv3time2 = rv.normal(0,1). compute dv3time3 = rv.normal(0,1). compute dv4time1 = rv.normal(0,1). compute dv4time2 = rv.normal(0,1). compute dv4time3 = rv.normal(0,1). end case. end loop. end file. end input program. execute. GLM dv1time1 dv1time2 dv1time3 dv2time1 dv2time2 dv2time3 dv3time1 dv3time2 dv3time3 dv4time1 dv4time2 dv4time3 BY group /WSFACTOR = Time 3 Polynomial /MEASURE = DV1 DV2 DV3 DV4 /METHOD = SSTYPE(3) /EMMEANS = TABLES(group) /EMMEANS = TABLES(Time) /EMMEANS = TABLES(group*Time) /PRINT = DESCRIPTIVE HOMOGENEITY /CRITERIA = ALPHA(.05) /WSDESIGN = Time /DESIGN = group . * Here is a more specific version of my question. * * An earlier "Multivariate Tests^a" output table includes the line: * Within Subjects Time Pillai's Trace 0.145 1.935^b 8.000 91.000 0.064 * * Two tables farther along is a "Multivariate^a,b" table that includes the line: * Within Subjects Time Pillai's Trace 0.068 1.712 8.000 388.000 0.094 * * These two tables include the same sources (i.e., Time and Time*group), * but they give different values and different error df's. * * What is the difference between these two tables? * (My question pertains to both of the sources, Time and Time*group).  View this message in context: http://spssxdiscussion.1045642.n5.nabble.com/QuestionaboutSPSSMANOVAoutputtp5733053.html Sent from the SPSSX Discussion mailing list archive at Nabble.com. ===================== To manage your subscription to SPSSXL, send a message to [hidden email] (not to SPSSXL), with no body text except the command. To leave the list, send the command SIGNOFF SPSSXL For a list of commands to manage subscriptions, send the command INFO REFCARD ________________________________________ Professor Diana Kornbrot Work University of Hertfordshire College Lane, Hatfield, Hertfordshire AL10 9AB, UK +44 (0) 170 728 4626 [hidden email] http://dianakornbrot.wordpress.com/ http://go.herts.ac.uk/Diana_Kornbrot skype: kornbrotme Home 19 Elmhurst Avenue London N2 0LT, UK +44 (0) 208 444 2081 ===================== To manage your subscription to SPSSXL, send a message to [hidden email] (not to SPSSXL), with no body text except the command. To leave the list, send the command SIGNOFF SPSSXL For a list of commands to manage subscriptions, send the command INFO REFCARD ===================== To manage your subscription to SPSSXL, send a message to [hidden email] (not to SPSSXL), with no body text except the command. To leave the list, send the command SIGNOFF SPSSXL For a list of commands to manage subscriptions, send the command INFO REFCARD 
Administrator

The following MANOVA syntax appears to produce the same results as Jeff's GLM command:
MANOVA dv1time1 dv1time2 dv1time3 dv2time1 dv2time2 dv2time3 dv3time1 dv3time2 dv3time3 dv4time1 dv4time2 dv4time3 BY group(1,2) /WSFACTORS=Time(3) /MEASURE = DV1 DV2 DV3 DV4 /* /WSDESIGN=Time, Time BY Group */ /PRINT=SIGNIF(HYPOTH AVERF) /DESIGN=GROUP . I commented out the WSDESIGN subcommand, because it was causing the syntax to fail. While exploring that, I pasted the following example from the FM into my syntax editor, and key words MANOVA and WSDESIGN appeared in red (rather than blue & green respectively), indicating that SPSS didn't like the command syntax. Example from the FM: MANOVA TEMP1 TO TEMP6, WEIGHT1 TO WEIGHT6 BY GROUP(1,2) /WSFACTORS=DAY(3) AMPM(2) /MEASURE=TEMP WEIGHT /WSDESIGN=DAY, AMPM, DAY BY AMPM /PRINT=SIGNIF(HYPOTH AVERF) /DESIGN. Anyway, the output from my MANOVA syntax above can be viewed in the attached text file. (Nonnabble users will have to visit this thread in Nabble to get it.) For convenience, I pasted results from GLM right below the corresponding results from MANOVA. Inside the text file, search for <FROM GLM:> to find those pasted results. As you will see, the MANOVA command above has produced all the same multivariate tests. HTH. Miller_example_via_MANOVA.txt p.s.  Jon or Rick, or anyone else who might know: Is the WSDESIGN subcommand for MANOVA no longer supported?

Bruce Weaver bweaver@lakeheadu.ca 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 email, please use the address shown above. 
In reply to this post by Bruce Weaver
Bruce, that might be. I didn't open (didn't notice) the files
provided by the OP and probably assumed about the question
incorrectly.
06.09.2016 16:12, Bruce Weaver пишет:
That doesn't sound right to me, Kirill. As the attached Excel file shows, *both* tables Jeff refers to are showing multivariate test results for the Time and Time*Group terms (with no separation by DV1 through DV4). Members who do not use Nabble can download the attached Excel file here: http://spssxdiscussion.1045642.n5.nabble.com/QuestionaboutSPSSMANOVAoutputtd5733053.html Miller_example.xls <http://spssxdiscussion.1045642.n5.nabble.com/file/n5733058/Miller_example.xls>
===================== To manage your subscription to SPSSXL, send a message to [hidden email] (not to SPSSXL), with no body text except the command. To leave the list, send the command SIGNOFF SPSSXL For a list of commands to manage subscriptions, send the command INFO REFCARD 
[This is a followup from the OP.]
First, thanks to everyone who replied. I greatly appreciate your time. The responses gave me several new links and pointers that look helpful, and I will definitely follow those up. Perhaps I should say that I am asking this question mostly to inform myself so that I can better teach MANOVA to some PhD students—not because I want to analyze an actual data set. As we go through this SPSS output, students often ask seemingly obvious questions that I can’t answer. For example, students want to know what, exactly, is meant by the table note “b. Tests are based on averaged variables”? One naive interpretation of this note is that this table is based on an analysis of a “dvavg” that one could compute (in my example) with these statements: compute dvavgtime1=(dv1time1+dv2time1+dv3time1+dv4time1)/4. compute dvavgtime2=(dv1time2+dv2time2+dv3time2+dv4time2)/4. compute dvavgtime3=(dv1time3+dv2time3+dv3time3+dv4time3)/4. execute. If this interpretation were correct, then surely some GLM of the dvavgtimeN scores would give the same results shown in the earlier table marked with note b, but I can find no such analysis. Evidently, the variables are being averaged in some other way, and it would be nice to know how. Students also ask about the error df’s in the different tables. Why are these values on the order of 380 for the table marked with note b, as compared with error df’s of 91 for the same sources in the original multivariate table? And what is going on with Roy’s largest root, which has the same hypothesis & error df’s as the other tests in the original table but has only half as many df’s for hypothesis and error in the table marked with note b? These students have a good grasp of how df’s relate to observations and model terms. It would be great to be able to give them some conceptual feeling for the models from which these dfs emerge—something more concrete than my current vague mumbling about “univariate approach” versus “multivariate approach”. Finally, I notice that Mike Palij suggested that I don't really have a doubly multivariate situation but a "straight" univariate withinsubject design requiring a repeated measures ANOVA. I’m not sure how I gave that impression. If it was that I used rv.normal(0,1) for all four dv’s then I apologize for being unclear—probably I should have used very distinct mu’s and sigma’s for the four dv’s so that it would be obvious that they were incommensurable quantities—i.e., that I really had a doubly multivariate design. 
Administrator

Go to Help Algorithms.
Should address all of these questions and more. 
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?" 
> Go to Help Algorithms.
> Should address all of these questions and more.  Thanks for the tip, but unfortunately this material doesn't help me at all. A big problem is that I can't see how to connect the algorithms described in the Help to the two output tables that I am trying to understand. For example, a single general model is specified under "Model (GLM algorithms)". I guess both tables are computed using versions of this model, but there is no indication understandable by me (yes, I realize this may be the problem) of how that model is applied differently in these two tables (e.g., constrained in one table). 
The What's This help might provide useful information here, though I haven't looked at the specific text. To see it, double click the table and then right click on a column header. On Tue, Sep 13, 2016 at 10:17 PM, Jeff Miller <[hidden email]> wrote: > Go to Help Algorithms. 
In reply to this post by Jeff Miller
Previously Dave Marso wrote:
>> Go to Help Algorithms. >> Should address all of these questions and more. > On Wednesday, September 14, 2016 12:17 AM, Jeff Miller wrote: > Thanks for the tip, but unfortunately this material doesn't help >me at all. A big problem is that I can't see how to connect the >algorithms described in the Help to the two output tables that I >am trying to understand. It is not clear what it is that you don't understand. Let's be more specific: I don't know what version of SPSS you're using but since I'm using ver 23, let's work with that. Let me suggest that you use the PDF for the Algorithms (it should be located in the directiory for IBM SPSS); if you can't find on your machine, go to the IBM SPSS webiste and download a copy; go to: http://www01.ibm.com/support/docview.wss?uid=swg27043946#en If you go to page 492, it will put you on a page for GLM algorithms. for the two version of the multivariate tests. The top part of the page describing the calculationss for the Multivariate tests for WihtinSubjects without the assumption of multivariate sphericity  Steps 2 and 3 specifies how the S matrice (for Hypothesis and Error matrices) are computefd for each of the multivariate tests (see page 482484 for the multivariate test calculations). This is done for each level of the betweensubjects variable. The bottom of the page covers the "Averaged Tests on the Within Subjects Effects" (aka Scheffe's Mixed Multivariate Model). Steps 1 to 6 identifies the specific calculations used and you can compare them to the calculations used in steps 2 and 3 in the Multivariate tests above. Is the problem you can't follow the matrix operations specified in the two sections or is it something else? You may want to look at the following references: The original article by Scheffe describing his mixed model analysis: Scheffe, H. (1956). A" mixed model" for the analysis of variance. The Annals of Mathematical Statistics, 2336. A PDF copy is available on the Project Euclid website; see: http://projecteuclid.org/euclid.aoms/1177728348 I believe I previously mentioned the following article by Boik but repeat it here for completeness sake: Boik, R. J. (1991). Scheffés mixed model for multivariate repeated measures: a relative efficiency evaluation. Communications in StatisticsTheory and Methods, 20(4), 12331255. Availabe at: http://www.tandfonline.com/doi/pdf/10.1080/03610929108830562?needAccess=true And you might want to look at the following article by Davidson that compares the univariate repeated measures ANOVA to the Averaged Multivariate/Scheffe MMM analysis (when the assumptions for the Average Multivariate test are met, it is more powerful than the univariate ANOVA and the nonaverage MANOVA  the problem is that SPSS does not provide a test for multivariate sphericity, etc.); see: Davidson, M. L. (1972). Univariate versus multivariate tests in repeatedmeasures experiments. Psychological Bulletin, 77(6), 446. It seems to me that SPSS promotes the use of the nonaveraged MANOVA for withinsubject variables even when they are all the same (e.g., reaction times under different conditions). If you have a copy of Tabachnick & Fidell (6th Edition), see pages 310311 for details on doing MANOVA/MANCOVA in SPSS using either the procedure MANOVA or GLM  they differ in what they provide and depending upon one's need, one might prefer one over the other or use both. >For example, a single general model is specified under "Model (GLM >algorithms)". > I guess both tables are computed using versions of this model, but > there is > no indication understandable by me (yes, I realize this may be the > problem) > of how that model is applied differently in these two tables (e.g., > constrained in one table). Mike Palij New York University [hidden email] ===================== To manage your subscription to SPSSXL, send a message to [hidden email] (not to SPSSXL), with no body text except the command. To leave the list, send the command SIGNOFF SPSSXL For a list of commands to manage subscriptions, send the command INFO REFCARD 
Thanks very much, Mike. The detailed road map you provided should help me a lot.
> Is the problem you can't follow the matrix operations specified in the two sections Well, yes and no. I believe that I could reproduce the algorithm (say, in MATLAB), so I can follow it in that sense. At this point I would be very hard pressed to say what was going on, though. I find step 5 under the Averaged Tests to be the most mysterious. It appears to me that the averaging takes place here, but I am not sure. Maybe this will become clear when I work through some examples. I will check out all of the references you suggested; so far, I’ve only looked at this one: > Davidson, M. L. (1972). Univariate versus multivariate tests in repeatedmeasures experiments. Initially, this only seems to compare the GG adjusted univariate versus multivariate tests in cases where there is only one DV (e.g., RT in a variety of conditions tested within Ss)—not the doubly repeatedmeasures case. But I believe the regular multivariate versus “averaged tests" multivariate analyses that I am struggling with are most distinct in the doubly repeatedmeasures case. I’ll have to study the article a bit more—maybe I am just overlooking the “averaged tests” bit. > It is not clear what it is that you don't understand. Such a broad topic…For starters, I would like to be able to explain to students: 1) What is the difference between the conclusions that would be appropriate from significant effects in the regular multivariate versus “averaged tests” multivariate? 2) Under what circumstances* would one test have more power than the other? (*For example, it would be nice to be able to generate simulated data favoring one test or the other.) 
Free forum by Nabble  Edit this page 