History of AFNI updates  

|
July 02, 2009 02:03PM
Jatin,

The conventional approach to testing linear trend with 3 levels, assuming equally spaced, is to assign weights of -1, 0, and 1. If you are concerned that the middle level is not really considered in the test, you can run two extra tests, one with High vs. Medium, and the other with Medium vs. Low. The combination of the 3 tests should be good enough to identify linearity. In the meantime, add a quadratic test with weights of 1, -2, and 1, and see if quadratic relation exists among the 3 levels.

Gang
Subject Author Posted

linear effects

Jatin July 01, 2009 11:22PM

Re: linear effects

Gang Chen July 02, 2009 02:03PM

Re: linear effects

Jatin July 02, 2009 02:20PM

Re: linear effects

Gang Chen July 02, 2009 02:25PM