Syntax to run analysis of simple main effects
Locked
 
|
Hello,
I need to conduct an analysis of simple main effects. From what I can gather I can only do this with SPSS using syntax instead of the dialogue boxes. I ran a 2X2X2X2 ANOVA initially.
Any thoughts on the syntax I would use?
Thanks in advance,
Chris
|
Re: Syntax to run analysis of simple main effects
|
|
Read up on the LMATRIX subcommands in the
GLM procedure. If any of your variables in the 2X2X2X2
are repeated measures, you will need to utilize the MMATRIX subcommands also. Joe Burleson From: SPSSX(r)
Discussion [mailto:[hidden email]] On
Behalf Of C Finello Hello, I need to conduct an analysis of simple main effects. From what I can
gather I can only do this with SPSS using syntax instead of the dialogue
boxes. I ran a 2X2X2X2 ANOVA initially. Any thoughts on the syntax I would use? Thanks in advance, Chris |
|
|
In reply to this post by C Finello
here is how this works using the old MANOVA code. the
last \design statement is key. i believe that GLM is very
similar:
GLM DV by A(1,2) B(1,2)
/print cellinfo(means)
/Design = A within B(1) A within
B(2)
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of C Finello Sent: Monday, February 09, 2009 12:25 PM To: [hidden email] Subject: Syntax to run analysis of simple main effects Hello,
I need to conduct an analysis of simple main effects. From what I can
gather I can only do this with SPSS using syntax instead of the dialogue
boxes. I ran a 2X2X2X2 ANOVA initially.
Any thoughts on the syntax I would use?
Thanks in advance,
Chris
|
Re: Syntax to run analysis of simple main effects
|
|
In reply to this post by C Finello
I guess my question is why do simple main effects on a 2x2x2x2
design. Every result has only one degree of freedom and is testing exactly one
hypothesis. SO doing simple main effects strikes me as similar to doing
posttest comparison for a t test. Paul R. Swank, Ph.D Professor and Director of Research Children's Learning Institute University of Texas Health Science Center Houston, TX 77038 From: SPSSX(r) Discussion
[mailto:[hidden email]] On Behalf Of C Finello Hello, I need to conduct an analysis of simple main effects. From
what I can gather I can only do this with SPSS using syntax instead of the
dialogue boxes. I ran a 2X2X2X2 ANOVA initially. Any thoughts on the syntax I would use? Thanks in advance, Chris |
Re: Syntax to run analysis of simple main effects
|
|
Simple main effects are, indeed, like
doing t-tests within the ANOVA structure. If, for example, you have only a 2 X
2 ANOVA, both between-subjects variables, and you find a significant 2 X 2
interaction, then looking at the effects of variable B within the 2 levels of
variable A is an a priori contrast. Maybe variable A is gender, and variable B
an experimental treatment. If, for example, there is no main effect for
variable B, but a significant A X B interaction, maybe the simple main effects
will show that the Tx is significantly effective for boys but non-sig for girls.
But it might also show, instead, that the Tx is significant in the wrong
direction for girls. You may wish to know these “simple effects”
after the fact. If you have a more complex design, a 2 X 2
X 2 X 2, and you get a sig 4-way interaction, then you may wish to explore the “simple”
3-way interaction effects within each of the levels of the 4th variable.
Again, if you got a Tx X Gender X Age X Ethnicity effect, maybe looking within
levels of Gender, you saw a significant 3-way Tx X Age X Ethnicity effect for
boys, but not a sig 3-way effect for girls. In this case, the “simple”
effects are no longer analogous to t-tests. Simple effects “cheat” in that
they take a 1 df test, as you point out, and do 2 analyses! Each effect is not
only “post-hoc”, but is underpowered due to the reduced sample
sizes. Still, too many researchers “block”
on one variable (i.e., gender) WITHOUT doing the interaction effect and then
report the two simple mains straight off. Not good. Joe Burleson From: SPSSX(r)
Discussion [mailto:[hidden email]] On
Behalf Of Swank, Paul R I guess my question
is why do simple main effects on a 2x2x2x2 design. Every result has only one
degree of freedom and is testing exactly one hypothesis. SO doing simple main
effects strikes me as similar to doing posttest comparison for a t test. Paul R. Swank, Ph.D Professor and
Director of Research Children's Learning
Institute University of From: SPSSX(r)
Discussion [mailto:[hidden email]] On
Behalf Of C Finello Hello, I need to conduct an analysis of simple main effects. From what I can
gather I can only do this with SPSS using syntax instead of the dialogue
boxes. I ran a 2X2X2X2 ANOVA initially. Any thoughts on the syntax I would use? Thanks in advance, Chris |
Re: Syntax to run analysis of simple main effects
|
|
You also find in a 2x2 design with a significant interaction
that neither group difference significant varies from zero, but they could
still differ significantly from each other. Paul R. Swank, Ph.D Professor and Director of Research Children's Learning Institute University of Texas Health Science Center Houston, TX 77038 From: Burleson,Joseph A.
[mailto:[hidden email]] Simple main effects are, indeed, like doing t-tests within the
ANOVA structure. If, for example, you have only a 2 X 2 ANOVA, both
between-subjects variables, and you find a significant 2 X 2 interaction, then looking
at the effects of variable B within the 2 levels of variable A is an a priori
contrast. Maybe variable A is gender, and variable B an experimental treatment.
If, for example, there is no main effect for variable B, but a significant A X
B interaction, maybe the simple main effects will show that the Tx is
significantly effective for boys but non-sig for girls. But it might also show,
instead, that the Tx is significant in the wrong direction for girls. You may
wish to know these “simple effects” after the fact. If you have a more complex design, a 2 X 2 X 2 X 2, and you get a
sig 4-way interaction, then you may wish to explore the “simple” 3-way
interaction effects within each of the levels of the 4th variable. Again,
if you got a Tx X Gender X Age X Ethnicity effect, maybe looking within levels
of Gender, you saw a significant 3-way Tx X Age X Ethnicity effect for boys,
but not a sig 3-way effect for girls. In this case, the “simple” effects are no
longer analogous to t-tests. Simple effects “cheat” in that they take a 1 df test, as you point
out, and do 2 analyses! Each effect is not only “post-hoc”, but is underpowered
due to the reduced sample sizes. Still, too many researchers “block” on one variable (i.e., gender)
WITHOUT doing the interaction effect and then report the two simple mains
straight off. Not good. Joe Burleson From: SPSSX(r) Discussion
[mailto:[hidden email]] On Behalf Of Swank, Paul R I guess my question is why do simple main effects on a 2x2x2x2
design. Every result has only one degree of freedom and is testing exactly one
hypothesis. SO doing simple main effects strikes me as similar to doing
posttest comparison for a t test. Paul R. Swank, Ph.D Professor and Director of Research Children's Learning Institute University of Texas Health Science Center Houston, TX 77038 From: SPSSX(r) Discussion
[mailto:[hidden email]] On Behalf Of C Finello Hello, I need to conduct an analysis of simple main effects. From
what I can gather I can only do this with SPSS using syntax instead of the
dialogue boxes. I ran a 2X2X2X2 ANOVA initially. Any thoughts on the syntax I would use? Thanks in advance, Chris |
Re: Syntax to run analysis of simple main effects
|
|
Indeed, this happens as often as not!
Another reason to at least know the individual, simple effects. Joe Burleson ________________________________ From: SPSSX(r) Discussion on behalf of Swank, Paul R Sent: Mon 2/9/2009 5:04 PM To: [hidden email] Subject: Re: Syntax to run analysis of simple main effects You also find in a 2x2 design with a significant interaction that neither group difference significant varies from zero, but they could still differ significantly from each other. Paul R. Swank, Ph.D Professor and Director of Research Children's Learning Institute University of Texas Health Science Center Houston, TX 77038 From: Burleson,Joseph A. [mailto:[hidden email]] Sent: Monday, February 09, 2009 3:41 PM To: Swank, Paul R; [hidden email] Subject: RE: Re: Syntax to run analysis of simple main effects Simple main effects are, indeed, like doing t-tests within the ANOVA structure. If, for example, you have only a 2 X 2 ANOVA, both between-subjects variables, and you find a significant 2 X 2 interaction, then looking at the effects of variable B within the 2 levels of variable A is an a priori contrast. Maybe variable A is gender, and variable B an experimental treatment. If, for example, there is no main effect for variable B, but a significant A X B interaction, maybe the simple main effects will show that the Tx is significantly effective for boys but non-sig for girls. But it might also show, instead, that the Tx is significant in the wrong direction for girls. You may wish to know these "simple effects" after the fact. If you have a more complex design, a 2 X 2 X 2 X 2, and you get a sig 4-way interaction, then you may wish to explore the "simple" 3-way interaction effects within each of the levels of the 4th variable. Again, if you got a Tx X Gender X Age X Ethnicity effect, maybe looking within levels of Gender, you saw a significant 3-way Tx X Age X Ethnicity effect for boys, but not a sig 3-way effect for girls. In this case, the "simple" effects are no longer analogous to t-tests. Simple effects "cheat" in that they take a 1 df test, as you point out, and do 2 analyses! Each effect is not only "post-hoc", but is underpowered due to the reduced sample sizes. Still, too many researchers "block" on one variable (i.e., gender) WITHOUT doing the interaction effect and then report the two simple mains straight off. Not good. Joe Burleson ________________________________ From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Swank, Paul R Sent: Monday, February 09, 2009 4:06 PM To: [hidden email] Subject: Re: Syntax to run analysis of simple main effects I guess my question is why do simple main effects on a 2x2x2x2 design. Every result has only one degree of freedom and is testing exactly one hypothesis. SO doing simple main effects strikes me as similar to doing posttest comparison for a t test. Paul R. Swank, Ph.D Professor and Director of Research Children's Learning Institute University of Texas Health Science Center Houston, TX 77038 From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of C Finello Sent: Monday, February 09, 2009 2:25 PM To: [hidden email] Subject: Syntax to run analysis of simple main effects Hello, I need to conduct an analysis of simple main effects. From what I can gather I can only do this with SPSS using syntax instead of the dialogue boxes. I ran a 2X2X2X2 ANOVA initially. Any thoughts on the syntax I would use? Thanks in advance, Chris ===================== 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 |
Re: Syntax to run analysis of simple main effects
|
|
Take the following data. 1 1 7 1 1 8 1 1 9 1 1 8 1 2 10 1 2 10 1 2 9 1 2 11 2 1 7 2 1 8 2 1 7 2 1 6 2 2 5 2 2 6 2 2 5 2 2 6 3 1 4 3 1 5 3 1 4 3 1 3 3 2 2 3 2 3 3 2 4 3 2 3 4 1 3 4 1 6 4 1 3 4 1 3 4 2 1 4 2 2 4 2 2 4 2 2 There are four levels of the first variable, DEP, and 2 levels
for the second, drug. Run simple main effects. They can be done two different
ways, looking at level of DEP for each drug and, perhaps more commonly, looking
at Drug for each level of DEP. If you do so and sum up the simple main effects
in each set you determine that the 4 effects for drug within DEP have a sum of
squares of 213.875, which is the sum of squares for DEP + the SS for the
interaction. Or if you sum the SS for DEP within drug, they add to 22.5 which
is the same as the SS for drug plus the interaction. Note also that there are 4
effects of drug within levels of DEP which is the df for DEP plus the df for
the interaction. So the simple main effects do confound the main effects with
the interaction. Dr. Paul R. Swank, Professor and Director of Research Children's Learning Institute University of Texas Health Science Center-Houston From: Burleson,Joseph A.
[mailto:[hidden email]] I’m not sure what you mean by redundant. Simple main effects
are not redundant with the main effects. They are also not redundant with the
interaction: they explain the pattern of the interaction more fully. As you
point out, there might be multiple patterns, all under the umbrella of a given
interaction. If Var A significantly interacts with Var B, then look, for
example, at the effects of A within B:
Since all of these three possibilities fall under the identical,
significant interaction, without doing the simple main effects, one is not able
to describe the above with any statistical confidence. With the simple mains,
it becomes more clear. Joe From: Swank, Paul R
[mailto:[hidden email]] They are redundant so why do you need them. They confound the
main effect and the interaction anyway. Paul R. Swank, Ph.D Professor and Director of Research Children's Learning Institute University of Texas Health Science Center Houston, TX 77038 From: Burleson,Joseph A.
[mailto:[hidden email]] Indeed, this happens as often as not! Another reason to at least know the individual, simple effects. Joe
Burleson From: SPSSX(r) Discussion on behalf of Swank,
Paul R You also find in a 2x2 design with a significant interaction
that neither group difference significant varies from zero, but they could
still differ significantly from each other. Paul R. Swank, Ph.D Professor and Director of Research Children's Learning Institute University of Texas Health Science Center Houston, TX 77038 From: Burleson,Joseph A.
[mailto:[hidden email]] Simple main effects are, indeed, like doing t-tests within the
ANOVA structure. If, for example, you have only a 2 X 2 ANOVA, both
between-subjects variables, and you find a significant 2 X 2 interaction, then
looking at the effects of variable B within the 2 levels of variable A is an a
priori contrast. Maybe variable A is gender, and variable B an experimental
treatment. If, for example, there is no main effect for variable B, but a
significant A X B interaction, maybe the simple main effects will show that the
Tx is significantly effective for boys but non-sig for girls. But it might also
show, instead, that the Tx is significant in the wrong direction for girls. You
may wish to know these “simple effects” after the fact. If you have a more complex design, a 2 X 2 X 2 X 2, and you get a
sig 4-way interaction, then you may wish to explore the “simple”
3-way interaction effects within each of the levels of the 4th variable.
Again, if you got a Tx X Gender X Age X Ethnicity effect, maybe looking within
levels of Gender, you saw a significant 3-way Tx X Age X Ethnicity effect for
boys, but not a sig 3-way effect for girls. In this case, the
“simple” effects are no longer analogous to t-tests. Simple effects “cheat” in that they take a 1 df test,
as you point out, and do 2 analyses! Each effect is not only
“post-hoc”, but is underpowered due to the reduced sample sizes. Still, too many researchers “block” on one variable
(i.e., gender) WITHOUT doing the interaction effect and then report the two
simple mains straight off. Not good. Joe Burleson From: SPSSX(r) Discussion
[mailto:[hidden email]] On Behalf Of Swank, Paul R I guess my question is why do simple main effects on a 2x2x2x2
design. Every result has only one degree of freedom and is testing exactly one
hypothesis. SO doing simple main effects strikes me as similar to doing
posttest comparison for a t test. Paul R. Swank, Ph.D Professor and Director of Research Children's Learning Institute University of Texas Health Science Center Houston, TX 77038 From: SPSSX(r) Discussion
[mailto:[hidden email]] On Behalf Of C Finello Hello, I need to conduct an analysis of simple main effects. From
what I can gather I can only do this with SPSS using syntax instead of the
dialogue boxes. I ran a 2X2X2X2 ANOVA initially. Any thoughts on the syntax I would use? Thanks in advance, Chris |
| Free forum by Nabble | Edit this page |