redzils: (Default)
[personal profile] redzils
I think I have solved the outstanding nagging question for my dissertation defense prep: one of the people on my committee looked at my results section before Christmas, and wanted to know if changing my mediation-religion, essentially (Baron and Kenney, 1984 to McKinnon and colleagues, 2002) would have had me doing different analyses. The answer is, well, maybe: my high holy mathemeticians spell out the conditions under which it is appropriate to test mediation (which were not met, so I found another analysis), while hers' does not. There is no defensible version where I say, "oh your guy totally supports my assumptions" - he Just Never Says, Either Way.

So, I've been wondering for the last, oh, six weeks, on and off how I could prepare to get through that gracefully, and without doing Structural Equation Modeling (which is very time consuming and wouldn't have given me much). Tonight I looked at her guy's website again and realized there was just no way out. But, instead of giving in and spending the next four days doing SEM, I tackled the three paths we are quarreling about one at a time, via Sobel test. None of them were signficant, so my decision to not test the complete model was supported, regardless of the assumptions you are prepared to make.

Essentially: I win.

I still need to figure out distribution data for the measures I used when administered in other samples and compare it to what I found inmy sample, but this is not hard, just annoying. And then I will be as prepared as I can be, until I meet with my advisor tomorrow morning. I expect her to ask for a few changes and new thoughts, but am still really, really hoping to pretty much wrap my prep up by close of business tomorrow, so I can spend my weekend merely reviewing the paper and my notes, not frantically flailing.

Profile

redzils: (Default)
redzils

November 2010

S M T W T F S
 123456
78910111213
14151617181920
2122 2324252627
282930    

Most Popular Tags

Style Credit

Expand Cut Tags

No cut tags
Page generated Sep. 26th, 2017 02:46 pm
Powered by Dreamwidth Studios