Programming the statistical procedures from SAS

orthogonal polynomials and 3-way interaction

Reply
New Contributor
Posts: 3

orthogonal polynomials and 3-way interaction

I have three factors: A, B (either qualitative or categorical) and C (quantitative, 4 levels at the least).

I want to write ESTIMATE and/or CONTRAST statements for the ff:
a. each level of A, A1, A2, A3
b. each level of B, B1, B2
c. 1st-3rd degree polynomial for C, Linear, Quadratic and Cubic
d. 2-way interactions, i.e. A1-B1, ...,A3-B2, A1-Linear, ..., A3-Cubic, B1-Linear, ..., B2xCubic
e. 3-way interactions, i.e. (A1-B1 v A1-B2)-Linear, ..., (A3-B1 v A3-B2)-Cubic

Purpose are (1) to determine the appropriate degree for each AxB level and (2) test for significant differences among AxB polynomials.

I know how to assign the coefficients for items (a) to (d). Example for a two-way interaction the CONTRAST statement is this:

contrast 'A1 v A2 Linear on C' AxC
-0.467435 -0.367181 -0.273192 -0.191736 -0.060152 0.0275699 0.1403557 0.2782051 0.3659274 0.547638
0.467435 0.367181 0.273192 0.191736 0.060152 -0.0275699 -0.1403557 -0.2782051 -0.3659274 -0.547638;

(C has 10 levels and is unequally spaced. I used the ORPOL function to get these coefficients)

For the 3-way interaction, I'm using the same rule for the 2-way interaction, ex.
contrast 'Linear: A1-B1 v A1-B2 Linear' A*B*C
-0.467435 -0.367181 -0.273192 -0.191736 -0.060152 0.0275699 0.1403557 0.2782051 0.3659274 0.547638
0.467435 0.367181 0.273192 0.191736 0.060152 -0.0275699 -0.1403557 -0.2782051 -0.3659274 -0.547638;


For this 3-way contrast, I get the NON-EST and I don't know what am I doing wrong. Can anyone please shed light on this?
Ask a Question
Discussion stats
  • 0 replies
  • 101 views
  • 0 likes
  • 1 in conversation