Type III test and LSMeans in Proc GLM for ANCOVA model do not agree

[fdehkord]

Hello,

I am new to Proc GLM and would appreciate help answering the following question.

I have fitted an ANCOVA model with covariate X (continuous) and two factors A and B (each with two Levels). The relationship between Y (the continuous response variable) and X is not linear therefore I have included a quadratic term for X in the model and included the interaction term x*A*B in the model as well. 

The type III F-test is significant for Factor A but not significant for factor B. The interaction term X*A*B is statistically significant and x and x2 are also significant.

The LSMEANS test for Factor A is not significant while it is significant for factor B ( the opposite of Type III F-test).

Why is this?  

[PaigeMiller]

Show us the output from your PROC GLM. Use the "Insert Photos" icon to include a screen capture of your output in your reply — do not attach files.

[fdehkord]

[Rick_SAS]

Interpreting higher-order interactions can be a challenge if you have not included the lower-order interactions. Have you also included the second-order interactions x*A, x*B, and A*B?  Show use your GLM code.

[fdehkord]

Here it is

Thanks.

[fdehkord]

I did include the lower-order terms and still have the same problem.