Runing contrast tests in a 2x2x2

[itwardus]

I am having some serious difficulty with getting my contrast test to run in my 2x2x2 ANCOVA. However I keep getting this error: This L matrix is not estimable. Hypothesis tests cannot be computed.


I have tried to run both a Special Contrast and a LMatrix. Could someone help me set up the syntax that I would need to test between two conditions for DV1 by A B C WITH CV1?

Thanks in advance.

[Bruce Weaver]

It would help if you generated a small dataset to illustrate the problem, and showed the syntax for what you have tried so far.  

Also, given that it is a 2x2x2, I wonder if you can get what you want via EMMEANS with COMPARE.  There is an example in the Command Syntax Reference manual entry for UNIANOVA.  You may be able to modify it to get the contrast you want.  

https://www.ibm.com/support/knowledgecenter/es/SSLVMB_21.0.0/com.ibm.spss.statistics.help/syn_unianova_emmeans.htm

HTH.

itwardus wrote

I am having some serious difficulty with getting my contrast test to run in my 2x2x2 ANCOVA. However I keep getting this error: This L matrix is not estimable. Hypothesis tests cannot be computed.


I have tried to run both a Special Contrast and a LMatrix. Could someone help me set up the syntax that I would need to test between two conditions for DV1 by A B C WITH CV1?

Thanks in advance.

[Mike]

Hi, just a couple of points:

(1)  Bruce's response is the first message to make it to the
SPSS mailing list (unless my email had a temporary psychotic
break with reality).

(2)  It would be nice if the OP would specify which contrast(s)
need to be done.  If we use A, B, and C represent the independent
variables, then is it some contrast in a 2-way interaction or
the 3-way interaction that needs to be done?

(3)  I could be wrong but I think there are some errors on the
page Bruce links to, specifically in the syntax provided.
I don't use Unianova often (I generally use GLM), but
shouldn't the first line of syntax be:

Unianova DEP by A B with X

where X is the covariate?  That's how it's done in GLM and
apparently in Unianova if one goes by the examples on
this webpage:
https://www.ibm.com/support/knowledgecenter/es/SSLVMB_21.0.0/com.ibm.spss.statistics.help/syn_unianova_examples.htm

If this really error, I worry what else is incorrect in the info.
Just checkin.

-Mike Palij
New York University
[hidden email]



----- Original Message -----
From: "Bruce Weaver" <[hidden email]>
To: <[hidden email]>
Sent: Wednesday, August 16, 2017 4:11 PM
Subject: Re: Runing contrast tests in a 2x2x2

> It would help if you generated a small dataset to illustrate the
> problem, and
> showed the syntax for what you have tried so far.
>
> Also, given that it is a 2x2x2, I wonder if you can get what you want
> via
> EMMEANS with COMPARE.  There is an example in the Command Syntax
> Reference
> manual entry for UNIANOVA.  You may be able to modify it to get the
> contrast
> you want.
>
> https://www.ibm.com/support/knowledgecenter/es/SSLVMB_21.0.0/com.ibm.spss.statistics.help/syn_unianova_emmeans.htm
>
> HTH.
>
>
> itwardus wrote
>> I am having some serious difficulty with getting my contrast test to
>> run
>> in my 2x2x2 ANCOVA. However I keep getting this error: This L matrix
>> is
>> not estimable. Hypothesis tests cannot be computed.
>>
>>
>> I have tried to run both a Special Contrast and a LMatrix. Could
>> someone
>> help me set up the syntax that I would need to test between two
>> conditions
>> for DV1 by A B C WITH CV1?
>>
>> Thanks in advance.
>
>
>
>
>
> -----
> --
> 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 e-mail, please use the address shown above.
>
> --
> View this message in context:
> http://spssx-discussion.1045642.n5.nabble.com/Runing-contrast-tests-in-a-2x2x2-tp5734662p5734663.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

[itwardus]

Thanks for the quick replies, I apologize I am a bit of a syntax novice and have been pouring over websearches all day to no avail.

Each IV in the 2x2x2 has 2 levels so there are 8 total conditions.

I am trying to do a simple contrast of two conditions so for example, A1B0C0 vs. A1B1C0.

I attempted the syntax as :
GLM DV1 BY A B C WITH CV1
/DESIGN CV1 A B C A*B A*C B*C A*B*C
/INTERCEPT=EXCLUDE
/PRINT = TEST(LMATRIX)
/LMARTIX A*B 0 0 -1 1  A*B*C 0 0 0 0 1 0 -1 0.

I have tried a few other iterations of the LMATRIX but none seem to work and I am a bit off on where exactly I am going off the rails.

[Bruce Weaver]

I don't have SPSS on this machine, so cannot test.  But if you try this:

GLM DV1 BY A B C WITH CV1
 /DESIGN CV1 A B C A*B A*C B*C A*B*C
 /EMMEANS TABLES(A*B*C) WITH(CV1=MEAN) COMPARE(B) ADJ(LSD)
 /INTERCEPT=INCLUDE.

I expect one of the pairwise contrasts produced by COMPARE(B) will be the A1B0C0 vs. A1B1C0 contrast you want.  If so, this is much easier (IMO) than messing around with LMATRIX.

By the way, notice that I changed the syntax to include the intercept, because I cannot imagine why you were excluding it.  If you have some good reason for excluding the intercept, please explain.

p.s. - Mike, I agree that the syntax shown on that IBM website I referred to earlier should have included WITH covariate.  

itwardus wrote

Thanks for the quick replies, I apologize I am a bit of a syntax novice and have been pouring over websearches all day to no avail.

Each IV in the 2x2x2 has 2 levels so there are 8 total conditions.

I am trying to do a simple contrast of two conditions so for example, A1B0C0 vs. A1B1C0.

I attempted the syntax as :
GLM DV1 BY A B C WITH CV1
/DESIGN CV1 A B C A*B A*C B*C A*B*C
/INTERCEPT=EXCLUDE
/PRINT = TEST(LMATRIX)
/LMARTIX A*B 0 0 -1 1  A*B*C 0 0 0 0 1 0 -1 0.

I have tried a few other iterations of the LMATRIX but none seem to work and I am a bit off on where exactly I am going off the rails.

itwardus wrote

Thanks for the quick replies, I apologize I am a bit of a syntax novice and have been pouring over websearches all day to no avail.

Each IV in the 2x2x2 has 2 levels so there are 8 total conditions.

I am trying to do a simple contrast of two conditions so for example, A1B0C0 vs. A1B1C0.

I attempted the syntax as :
GLM DV1 BY A B C WITH CV1
/DESIGN CV1 A B C A*B A*C B*C A*B*C
/INTERCEPT=EXCLUDE
/PRINT = TEST(LMATRIX)
/LMARTIX A*B 0 0 -1 1  A*B*C 0 0 0 0 1 0 -1 0.

I have tried a few other iterations of the LMATRIX but none seem to work and I am a bit off on where exactly I am going off the rails.

[Ryan]

This issue has been discussed quite a bit in the past. You might consider searching the SPSS-L archives.

To make sure sure the coefficient matrix L is estimable, my general recommendation is to start by writing out the linear combinations to estimate cell means, and then to use those linear combinations to construct the contrasts of interest.

While I agree with Bruce that the EMMEANS statement is a simpler alternative, using the coefficient matrix L is very useful for educational purposes and constructing certain complex contrasts. If you are interested in learning more about this topic, write back with the current LMATRIX code you wrote, and time permitting we will help you arrive at estimable statements.

Ryan

Sent from my iPhone

> On Aug 16, 2017, at 3:12 PM, itwardus <[hidden email]> wrote:
>
> I am having some serious difficulty with getting my contrast test to run in
> my 2x2x2 ANCOVA. However I keep getting this error: This L matrix is not
> estimable. Hypothesis tests cannot be computed.    
>
>
> I have tried to run both a Special Contrast and a LMatrix. Could someone
> help me set up the syntax that I would need to test between two conditions
> for DV1 by A B C WITH CV1?
>
> Thanks in advance.
>
>
>
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Runing-contrast-tests-in-a-2x2x2-tp5734662.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