Repeated measures: Linear trend significant but not within-subjs test

[Allan Lundy, PhD]

Dear Listers,
I am running analyses of a number of different outcome measures to assess treatment over 7 months, with testing following each treatment.  N= 23.  No predictor variables other than the effect of time.  The client expects to see steady improvement over time, hence I am using GLM Repeated Measures with polynomial contrasts, and expecting significant linear effects.  On a few outcome measures, I am getting a significant linear trend (e.g., p= .034) but non-sig tests of within-subjects effect (e.g., p= .069).  Also, nowhere near sig multivariate tests, (p= .368), though I have read that that is irrelevant in such a simple analysis.  I assume that it would be conservative to trust the least significant result, but has anyone seen a discussion of when and why this occurs?  (I would have assumed that the within-subjects effect was an omnibus test, hence nothing under it would be significant.)  Has anyone seen an argument to the effect that the linear trend can be honestly significant even if the overall effect is not?

Thanks for your attention!
Allan

Allan Lundy, PhD
Research Consulting
[hidden email]

Business & Cell (any time): 215-820-8100
Home (8am-10pm, 7 days/week): 215-885-5313
Address:  108 Cliff Terrace, Wyncote, PA 19095
Visit my Web site at www.dissertationconsulting.net ===================== 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

[Swank, Paul R]

Polynomial contrasts are orthogonal. Since you planned these ahead of time, just look at the contrasts, not the omnibus tests.

 

Dr. Paul R. Swank,

Professor and Director of Research

Children's Learning Institute

University of Texas Health Science Center-Houston

 

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Allan Lundy, PhD
Sent: Wednesday, March 30, 2011 7:10 PM
To: [hidden email]
Subject: Repeated measures: Linear trend significant but not within-subjs test

 

Dear Listers,
I am running analyses of a number of different outcome measures to assess treatment over 7 months, with testing following each treatment.  N= 23.  No predictor variables other than the effect of time.  The client expects to see steady improvement over time, hence I am using GLM Repeated Measures with polynomial contrasts, and expecting significant linear effects.  On a few outcome measures, I am getting a significant linear trend (e.g., p= .034) but non-sig tests of within-subjects effect (e.g., p= .069).  Also, nowhere near sig multivariate tests, (p= .368), though I have read that that is irrelevant in such a simple analysis.  I assume that it would be conservative to trust the least significant result, but has anyone seen a discussion of when and why this occurs?  (I would have assumed that the within-subjects effect was an omnibus test, hence nothing under it would be significant.)  Has anyone seen an argument to the effect that the linear trend can be honestly significant even if the overall effect is not?

Thanks for your attention!
Allan

Allan Lundy, PhD
Research Consulting
[hidden email]

Business & Cell (any time): 215-820-8100
Home (8am-10pm, 7 days/week): 215-885-5313
Address:  108 Cliff Terrace, Wyncote, PA 19095
Visit my Web site at www.dissertationconsulting.net

===================== 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