Normally I don't like "top posting" (leaving several screensful of
quoted material at the bottom of one's post), but it seems expedient
here.

You've got repeated measures on a binary dependent variable, and what
that suggests to me is that you want to use either a GEE (generalized
estimating equation) or GLMM (generalized linear mixed model), with
food placement as a time-varying covariate.  SAS can handle either of
those analyses, or you can download free software that will do it
<http://tigger.uic.edu/~hedeker/mix.html>.

But I'm not sure whether you can do it with only one subject.

Anyone else have ideas?  Does this call for some sort of time-series
analysis?

--David Epstein
  [EMAIL PROTECTED]

On Thu, 26 May 2005, Don  Allen went:

Hi Tipsters-

I need some advice on the best statistic to use. I'm doing some work
that's outside of my normal area. I'm collaborating with a colleague in
the Biology department and we're investigating learning in Raptors. In
particular, red-tailed hawks. So far we have preliminary data on only one
bird. The initial paradigm is to see if we can get the bird to shift its
natural preference from feeding on the left side of the cage to the right
side. First, food was available on both sides, then less food was
available on the left and more was on the right. Finally, the bird only
found food on the right. The bird's behaviour was as follows:



-- Date         Side first chosen
May      6              L
        7               L
        8               L
        9               L
        10              L
        11              L, L (two trials)
        12              L, L

B)  One piece under L cup, three pieces under R cup.
May     13              L, L
        14              L, L
        15              L, L
        16              L, L

C) No food under L cup, four pieces under R cup.
May     17              L, L
        18              L, L
        19              L, L
        20              R, R
        21              R, L, R (three trials)
        22              R, R
        23              R, R, R

My question is which is the best statistic to use to show that this shift
in behaviour is beyond chance (alpha = .05) I seem to recall a program
that calculated odds in an ABAB design, but I can't seem to find it in the
Digest. Of course a simple conjoint probability calculation shows that the
odds of six rights in a row is well below .05 but is there a better test
to use?

As usual, thanks in advance for your help.

-Don


Don Allen
Langara College

---
You are currently subscribed to tips as: [email protected]
To unsubscribe send a blank email to [EMAIL PROTECTED]

Reply via email to