# Sequential SS in testing for main effects

1. Apr 18, 2013

### libragirl79

Linear Model topic:

Given that we have a crossed fixed model w/o interaction, how would I go about testing for the main effects using extra (seq) sum of squares?

I know I am supposed to somehow use the F test and get the non centrality parameter but I don't know how to start....

Thanks!!

2. Apr 18, 2013

### Ray Vickson

You need to supply much more information. I assume you have a 2-way layout, with $y_{ij}$ being the value for treatment i and column j. Do you have a rectangular matrix of data, or do you have different numbers of j's for different i's? Are the j's also 'treatments', or are they just repetitions of the same experiment (i.e., several measurements at the same i)?

Don't you have a textbook or course where all this material is discussed? Certainly, all this material is readily available on-line. Trying to learn it in a homework forum seems futile. We can supply hints only.

3. Apr 18, 2013

### libragirl79

It's all just theory, I don't have actually any numbers, i goes from 1 to a and j goes from 1 to b. It's yij=mu+alphai+bj+e. I need to somehow use extra sum of squares R(alpha given beta and mu) to get to the quadratic form which would have a non central distribution and divided by the MSE, would have non central F to test for the effect of alpha. My problem is that I don't know how get the quadratic form for the extra SS in this case....