Three-Way Anova Post Hoc Comparisons
Locked
 
|
In my data I have three independent variables; intervention (intervention vs control group), gender (male vs female) and country (country1, country2, country3). I have a three way interaction and I don't know how to conduct post hoc analyses. Basicly first I have to create a intervention-control difference. Then, I have to do this for gender specific. Afterwards I have to create a country specific difference for each gender specific intervention-control difference.
The most advanced post hoc comparison I have seen is 2- way mixed design until now. SO I don't know what to do with this 3-way post-hoc. I appreciate even a tiny bit of help:) Thank you very much. intervention control difference - intervention control difference for males for females @ country -1 @ country-2 |
Re: Three-Way Anova Post Hoc Comparisons
|
|
Try this (only available through syntax):
UNIANOVA Y BY A B C /METHOD=SSTYPE(3) /INTERCEPT=INCLUDE /EMMEANS=TABLES(A) COMPARE ADJ(LSD) /EMMEANS=TABLES(B) COMPARE ADJ(LSD) /EMMEANS=TABLES(C) COMPARE ADJ(LSD) /EMMEANS=TABLES(A*B) COMPARE (A) ADJ(LSD) /EMMEANS=TABLES(A*B) COMPARE (B) ADJ(LSD) /EMMEANS=TABLES(A*C) COMPARE (A) ADJ(LSD) /EMMEANS=TABLES(A*C) COMPARE (C) ADJ(LSD) /EMMEANS=TABLES(B*C) COMPARE (B) ADJ(LSD) /EMMEANS=TABLES(B*C) COMPARE (C) ADJ(LSD) /EMMEANS=TABLES(A*B*C) COMPARE (A) ADJ(LSD) /EMMEANS=TABLES(A*B*C) COMPARE (B) ADJ(LSD) /EMMEANS=TABLES(A*B*C) COMPARE (C) ADJ(LSD) /CRITERIA=ALPHA(.05) /DESIGN=A B C A*B A*C B*C A*B*C. Good luck! Best Regards, Joost van Ginkel Joost R. van Ginkel, PhD Leiden University Faculty of Social and Behavioural Sciences PO Box 9555 2300 RB Leiden The Netherlands Tel: +31-(0)71-527 3620 Fax: +31-(0)71-527 3945 -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of komakino0 Sent: Tuesday, January 11, 2011 11:19 AM To: [hidden email] Subject: Three-Way Anova Post Hoc Comparisons In my data I have three independent variables; intervention (intervention vs control group), gender (male vs female) and country (country1, country2, country3). I have a three way interaction and I don't know how to conduct post hoc analyses. Basicly first I have to create a intervention-control difference. Then, I have to do this for gender specific. Afterwards I have to create a country specific difference for each gender specific intervention-control difference. The most advanced post hoc comparison I have seen is 2- way mixed design until now. SO I don't know what to do with this 3-way post-hoc. I appreciate even a tiny bit of help:) Thank you very much. intervention control difference - intervention control difference for males for females @ country -1 @ country-2 -- View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Three-Way-Anova-Post-Hoc-C omparisons-tp3336188p3336188.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 |
|
|
Just a couple of comments:
(1) This type of analysis appears to straddle the fence between planned comparisons and post hoc multiple comparisons. For main effects, the distinction is not important but for the intereactions, they become relevant because of (a) Type I error concerns and (b) level of statistical power. (2) An ordinary ANOVA would produce 7 potentially significant effects (3 main effects and 4 interactions). One could argue that one can ignore the "family-wise Type I" error rate (which would be calculated by Overall alpha = (1 - (1-alpha per comparison)**7) Assuming a per comparisons/F test result is judged with a traditional per comparison alpha of .05, after evaluating the seven results, the overall Type I error rate is .30, that is, there is a 30% chance that one of the results is statistically significant. If one did all of the tests below, because there are 13 comparisons, the overall alpha would be .49. NOTE: since unadjusted or Lease Significant Difference (LSD) tests are being conducted and if the independent variables have more than 2 levels, there will be additional tests. It would appear that after all statistical tests are done, the overall probability of having made a Type I error would be around .50 or greater. (3) The main effects have more statistical power than the interactions and the lower order ----- Original Message ----- From: "Ginkel, Joost van" <[hidden email]> To: <[hidden email]> Sent: Tuesday, January 11, 2011 5:33 AM Subject: Re: Three-Way Anova Post Hoc Comparisons > Try this (only available through syntax): > > UNIANOVA Y BY A B C > /METHOD=SSTYPE(3) > /INTERCEPT=INCLUDE > /EMMEANS=TABLES(A) COMPARE ADJ(LSD) > /EMMEANS=TABLES(B) COMPARE ADJ(LSD) > /EMMEANS=TABLES(C) COMPARE ADJ(LSD) > /EMMEANS=TABLES(A*B) COMPARE (A) ADJ(LSD) > /EMMEANS=TABLES(A*B) COMPARE (B) ADJ(LSD) > /EMMEANS=TABLES(A*C) COMPARE (A) ADJ(LSD) > /EMMEANS=TABLES(A*C) COMPARE (C) ADJ(LSD) > /EMMEANS=TABLES(B*C) COMPARE (B) ADJ(LSD) > /EMMEANS=TABLES(B*C) COMPARE (C) ADJ(LSD) > /EMMEANS=TABLES(A*B*C) COMPARE (A) ADJ(LSD) > /EMMEANS=TABLES(A*B*C) COMPARE (B) ADJ(LSD) > /EMMEANS=TABLES(A*B*C) COMPARE (C) ADJ(LSD) > /CRITERIA=ALPHA(.05) > /DESIGN=A B C A*B A*C B*C A*B*C. > > Good luck! > > Best Regards, > > Joost van Ginkel > > Joost R. van Ginkel, PhD > Leiden University > Faculty of Social and Behavioural Sciences > PO Box 9555 > 2300 RB Leiden > The Netherlands > Tel: +31-(0)71-527 3620 > Fax: +31-(0)71-527 3945 > > > -----Original Message----- > From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of > komakino0 > Sent: Tuesday, January 11, 2011 11:19 AM > To: [hidden email] > Subject: Three-Way Anova Post Hoc Comparisons > > In my data I have three independent variables; intervention > (intervention vs control group), gender (male vs female) and country > (country1, country2, country3). I have a three way interaction and I > don't know how to conduct post hoc analyses. Basicly first I have to > create a intervention-control difference. Then, I have to do this for > gender specific. Afterwards I have to create a country specific > difference for each gender specific intervention-control difference. > The most advanced post hoc comparison I have seen is 2- way mixed design > until now. SO I don't know what to do with this 3-way post-hoc. I > appreciate even a tiny bit of help:) Thank you very much. > > > > intervention control difference - intervention control difference > for males for females > @ country -1 @ country-2 > -- > View this message in context: > http://spssx-discussion.1045642.n5.nabble.com/Three-Way-Anova-Post-Hoc-C > omparisons-tp3336188p3336188.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 ===================== 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 |
|
|
Thank you very much for your valuable comments and help. Actually I run this syntax to see which comparisons are relevant;
EMMEANS=TABLES(A*B*C) COMPARE (A) ADJ(LSD) This command gives me comparisons that I wanted. But about the type I error rate risk; what can I do? For instance would Bonferonni correction work here? Planned comparison might be even more preferable yes. Because I have a specific comparison in my mind. I am not trying to catch something significant. But this time I have to write L-Matrix commands for the contrast. As you can guess this would be much more challenging for me:) So, I am also open to suggestions on that part. |
Re: Three-Way Anova Post Hoc Comparisons
|
|
In reply to this post by komakino0
If you know the MANOVA procedure, I believe that it will alow you to specify everything you need.
Below is syntax for class example for 3-way design that should be helpful TITLE 3-WAY EXAMPLE; SEE MAXWELL & DELANEY PP.397-398. DATA LIST FREE /1 VALUE SLFMTR ARGMNT SOURCE. BEGIN DATA. 4 1 1 1 3 1 1 1 4 1 1 1 5 1 1 1 2 1 1 1 5 1 1 1 4 1 1 1 6 1 1 1 3 1 1 1 4 1 1 1 5 1 1 1 4 1 1 1 4 1 1 2 4 1 1 2 2 1 1 2 3 1 1 2 5 1 1 2 3 1 1 2 2 1 1 2 3 1 1 2 4 1 1 2 3 1 1 2 2 1 1 2 4 1 1 2 3 1 2 1 5 1 2 1 3 1 2 1 2 1 2 1 6 1 2 1 4 1 2 1 4 1 2 1 3 1 2 1 5 1 2 1 3 1 2 1 2 1 2 1 3 1 2 1 5 1 2 2 5 1 2 2 7 1 2 2 5 1 2 2 6 1 2 2 4 1 2 2 3 1 2 2 5 1 2 2 6 1 2 2 7 1 2 2 7 1 2 2 6 1 2 2 3 2 1 1 5 2 1 1 5 2 1 1 4 2 1 1 3 2 1 1 2 2 1 1 1 2 1 1 5 2 1 1 3 2 1 1 4 2 1 1 3 2 1 1 4 2 1 1 5 2 1 2 4 2 1 2 3 2 1 2 2 2 1 2 4 2 1 2 6 2 1 2 2 2 1 2 4 2 1 2 4 2 1 2 3 2 1 2 4 2 1 2 3 2 1 2 5 2 2 1 6 2 2 1 4 2 2 1 7 2 2 1 6 2 2 1 7 2 2 1 5 2 2 1 6 2 2 1 4 2 2 1 6 2 2 1 7 2 2 1 5 2 2 1 6 2 2 2 4 2 2 2 4 2 2 2 2 2 2 2 4 2 2 2 5 2 2 2 4 2 2 2 3 2 2 2 4 2 2 2 2 2 2 2 3 2 2 2 4 2 2 2 END DATA. VALUE LABEL SLFMTR 1'HI Slf-M' 2'LO Slf-M' /ARGMNT 1'STRONG' 2'WEAK' /SOURCE 1'EXPERT' 2'ATTRACTV'. GRAPH/LINE(MULTIPLE)MEAN(VALUE) BY ARGMNT BY SOURCE/INTERVAL SE(1). GRAPH/LINE(MULTIPLE)MEAN(VALUE) BY SLFMTR BY SOURCE/INTERVAL SE(1). GRAPH/LINE(MULTIPLE)MEAN(VALUE) BY ARGMNT BY SLFMTR/INTERVAL SE(1). GRAPH/LINE(MULTIPLE)MEAN(VALUE) BY ARGMNT BY SOURCE/PANEL COLVAR=SLFMTR /INTERVAL SE(1). *QUESTION B: omnibus factorial, see SOURCE x ARGMNT interaction. MANOVA VALUE BY SOURCE(1,2) ARGMNT(1,2) SLFMTR(1,2) /PRINT=CELLINFO(MEANS) /OMEANS=TABLES (SOURCE ARGMNT SLFMTR SOURCE*ARGMNT SOURCE*SLFMTR SLFMTR*ARGMNT) /DESIGN. *QUESTIONS C & D: simple interactions (SxA) @ EACH LEVEL OF SLFMTR. MANOVA VALUE BY SOURCE(1,2) ARGMNT(1,2) SLFMTR(1,2) /DESIGN=SLFMTR SOURCE ARGMNT SLFMTR*SOURCE SLFMTR*ARGMNT SOURCE*ARGMNT WITHIN SLFMTR(1) SOURCE*ARGMNT WITHIN SLFMTR(2). *QUESTIONS F thru I: simiple-simple main effects of ARGMNT. MANOVA VALUE BY SOURCE(1,2) ARGMNT(1,2) SLFMTR(1,2) /DESIGN=SLFMTR SOURCE SLFMTR*SOURCE ARGMNT WITHIN SLFMTR(1) WITHIN SOURCE(1) ARGMNT WITHIN SLFMTR(1) WITHIN SOURCE(2) ARGMNT WITHIN SLFMTR(2) WITHIN SOURCE(1) ARGMNT WITHIN SLFMTR(2) WITHIN SOURCE(2). _____________________________________________________________ Jason W. Beckstead, Ph.D. Associate Professor/Quantitative Methodologist University of South Florida College of Nursing 12901 Bruce B. Downs Blvd., MDC22, Tampa, FL 33612, USA Statistical Editor, International Journal of Nursing Studies phone: +1.813.974.7667 fax: +1.813.974.5418 personal website: http://personal.health.usf.edu/jbeckste/ International Journal of Nursing Studies http://www.elsevier.com/ijns -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of komakino0 Sent: Tuesday, January 11, 2011 5:19 AM To: [hidden email] Subject: Three-Way Anova Post Hoc Comparisons In my data I have three independent variables; intervention (intervention vs control group), gender (male vs female) and country (country1, country2, country3). I have a three way interaction and I don't know how to conduct post hoc analyses. Basicly first I have to create a intervention-control difference. Then, I have to do this for gender specific. Afterwards I have to create a country specific difference for each gender specific intervention-control difference. The most advanced post hoc comparison I have seen is 2- way mixed design until now. SO I don't know what to do with this 3-way post-hoc. I appreciate even a tiny bit of help:) Thank you very much. intervention control difference - intervention control difference for males for females @ country -1 @ country-2 -- View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Three-Way-Anova-Post-Hoc-Comparisons-tp3336188p3336188.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 |
|
|
In reply to this post by komakino0
I'm just getting around to this thread. One approach would be to
invoke the TEST subcommands in the MIXED procedure. I provide a couple of examples of contrasts. (Note that I start by generating data) Ryan -- *Generate data. set seed 4836235. new file. inp pro. comp personID = -99. comp gender = -99. comp condition = -99. comp error = -99. comp country1 = -99. comp country2 = -99. comp residual_variance = 0.05. leave personID to residual_variance. loop personID = 1 to 12000. comp gender = rv.bernoulli(0.50). comp condition = rv.bernoulli(0.50). if (personID <= 4000) country1 = 1. if (personID > 4000) country1 = 0. if (personID <= 4000) country2 = 0. if (personID > 4000 and personID <= 8000) country2 = 1. if (personID > 8000) country2 = 0. comp error = sqrt(residual_variance)*rv.normal(0,1). comp y = 4.0 + 1.0*condition + 1.5*country1 + 2.0*country2 + 2.5*gender + 3.0*condition*country1 + 3.5*condition*country2 + 4.0*condition*gender + 4.5*country1*gender + 5.0*country2*gender + 5.5*condition*country1*gender + 6.0*condition*country2*gender + error. end case. end loop. end file. end inp pro. exe. *Construct country indicator variable. if (country1=1) country=1. if (country2=1) country=2. if (country1 = 0 and country2 = 0) country = 3. exe. *delete irrelevant variables created in simulation. delete variables error residual_variance country1 country2. mixed y by condition country with gender /fixed=condition country gender condition*country condition*gender country*gender condition*country*gender | sstype(3) /print=solution lmatrix /test 'cond1 vs cond2|cntry=1,gender=1 VS cond1 vs cond2|cntry=2,gender=1' condition*country 1 -1 0 -1 1 0 condition*country*gender 1 -1 0 -1 1 0 /test 'cond1 vs cond2|cntry=1,gender=0 VS cond1 vs cond2|cntry=2,gender=0' condition*country 1 -1 0 -1 1 0 condition*country*gender 0 0 0 0 0 0. On Tue, Jan 11, 2011 at 5:19 AM, komakino0 <[hidden email]> wrote: > In my data I have three independent variables; intervention (intervention vs > control group), gender (male vs female) and country (country1, country2, > country3). I have a three way interaction and I don't know how to conduct > post hoc analyses. Basicly first I have to create a intervention-control > difference. Then, I have to do this for gender specific. Afterwards I have > to create a country specific difference for each gender specific > intervention-control difference. > The most advanced post hoc comparison I have seen is 2- way mixed design > until now. SO I don't know what to do with this 3-way post-hoc. I appreciate > even a tiny bit of help:) Thank you very much. > > > > intervention control difference - intervention control difference > for males for females > @ country -1 @ country-2 > -- > View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Three-Way-Anova-Post-Hoc-Comparisons-tp3336188p3336188.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 |