Analysis of crossover design

[Steve-41]

Hi All,

I've been given some data to analyze that compares two different teaching
techniques.  The study was a crossover design and the gist of the design was
this: Students were divided into groups A and B, and a pre-test was
performed to assess their pre-course knowledge.  Group A was then taught 6
different modules using teaching technique Y and group B was taught the same
modules using teaching technique Z.  Then a crossover was performed and 6
more modules (different than the previous ones) were taught to group A using
teaching technique Z and group B using teaching technique Y.  Finally a
post-test was performed on all students to assess post-course knowledge.

Both the pre and post-course knowledge assessments were performed using
tests that included equal questions from each of the 12 modules.

My questions is this: What is the best way to set up the data for the
following analysis?

I was planning on analyzing the data using ANCOVA. My post-test scores would
be the dependent variable, pre-test would be a covariate and the teaching
technique would be the factor. However, this is where I get stuck.  Should I
break the test scores into two separate scores (one score for the first 6
modules and another score for the latter 6 modules)?

I feel that leaving the pre- and post-test scores as they are will fail to
account for the fact that some things were taught one way while others were
taught another way.  I'm not sure if the way SPSS handles ANCOVA (or ANOVA)
would account for that or would I need some modification of the dataset
prior to analysis.

I hope I explained this clearly enough, since it's confusing me more the
more I think of it.  Any help would be greatly appreciated. Thanks.

- Steve

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

[Reutter, Alex]

Hi Steve,

Have a look at: Help > Case Studies, then Advanced > Linear Mixed Models > Using Linear Mixed Models to Analyze a Crossover Trial.  It's not exactly the same setup that you have, but should give you some ideas.

Cheers,
Alex

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Steve
Sent: Wednesday, March 11, 2009 2:51 PM
To: [hidden email]
Subject: Analysis of crossover design

Hi All,

I've been given some data to analyze that compares two different teaching
techniques.  The study was a crossover design and the gist of the design was
this: Students were divided into groups A and B, and a pre-test was
performed to assess their pre-course knowledge.  Group A was then taught 6
different modules using teaching technique Y and group B was taught the same
modules using teaching technique Z.  Then a crossover was performed and 6
more modules (different than the previous ones) were taught to group A using
teaching technique Z and group B using teaching technique Y.  Finally a
post-test was performed on all students to assess post-course knowledge.

Both the pre and post-course knowledge assessments were performed using
tests that included equal questions from each of the 12 modules.

My questions is this: What is the best way to set up the data for the
following analysis?

I was planning on analyzing the data using ANCOVA. My post-test scores would
be the dependent variable, pre-test would be a covariate and the teaching
technique would be the factor. However, this is where I get stuck.  Should I
break the test scores into two separate scores (one score for the first 6
modules and another score for the latter 6 modules)?

I feel that leaving the pre- and post-test scores as they are will fail to
account for the fact that some things were taught one way while others were
taught another way.  I'm not sure if the way SPSS handles ANCOVA (or ANOVA)
would account for that or would I need some modification of the dataset
prior to analysis.

I hope I explained this clearly enough, since it's confusing me more the
more I think of it.  Any help would be greatly appreciated. Thanks.

- Steve

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

[Burleson,Joseph A.]

Code Orders A and B as a variable (1,2), then include as a
between-subjects variable.

Insofar as the statistic: presumably you have a continuous measure of
knowledge. Assuming it is normally distributed (transforms might be
necessary if it is not), consider doing a 2 (within: Time) X 2 (within:
Technique A versus b) X 2 (Order) repeated-measures ANOVA, with all
interactions included (Time X Technique, Time X Order, & Time X
Technique X Order). Use a sequential model.

Joe Burleson

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Steve
Sent: Wednesday, March 11, 2009 3:51 PM
To: [hidden email]
Subject: Analysis of crossover design

Hi All,

I've been given some data to analyze that compares two different
teaching
techniques.  The study was a crossover design and the gist of the design
was
this: Students were divided into groups A and B, and a pre-test was
performed to assess their pre-course knowledge.  Group A was then taught
6
different modules using teaching technique Y and group B was taught the
same
modules using teaching technique Z.  Then a crossover was performed and
6
more modules (different than the previous ones) were taught to group A
using
teaching technique Z and group B using teaching technique Y.  Finally a
post-test was performed on all students to assess post-course knowledge.

Both the pre and post-course knowledge assessments were performed using
tests that included equal questions from each of the 12 modules.

My questions is this: What is the best way to set up the data for the
following analysis?

I was planning on analyzing the data using ANCOVA. My post-test scores
would
be the dependent variable, pre-test would be a covariate and the
teaching
technique would be the factor. However, this is where I get stuck.
Should I
break the test scores into two separate scores (one score for the first
6
modules and another score for the latter 6 modules)?

I feel that leaving the pre- and post-test scores as they are will fail
to
account for the fact that some things were taught one way while others
were
taught another way.  I'm not sure if the way SPSS handles ANCOVA (or
ANOVA)
would account for that or would I need some modification of the dataset
prior to analysis.

I hope I explained this clearly enough, since it's confusing me more the
more I think of it.  Any help would be greatly appreciated. Thanks.

- Steve

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

[Steve-41]

Alex,

Thanks for the hint.  I checked out the case study and it definitely
helped. :)

- Steve



On Wed, 11 Mar 2009 15:44:41 -0500, Reutter, Alex <[hidden email]>
wrote:

>Hi Steve,
>
>Have a look at: Help > Case Studies, then Advanced > Linear Mixed Models
> Using Linear Mixed Models to Analyze a Crossover Trial.  It's not
exactly the same setup that you have, but should give you some ideas.
>
>Cheers,
>Alex
>

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