Eugene Tunik wrote:
> I have 4 conditions (A, B, C, D). I want to look at linear increases
> across conditions B, C, D only.
> Specifying a contrast vector of [0 -1 0 1] is clearly wrong.
> Specifying a vector of [0 1 2 3] is probably wrong. I know that
> weights have to add to zero generally -- though perhaps this doesn't
> apply when a when a linear trend is hypothesized???? I don't know...
> In reading this chapter (attached, pg 9) I see that this vector would
> not give me an exclusively linear trend.
>
> I guess I can do an F contrast with weights:
> [0 -1 1 0,
> 0 0 -1 1]
> but the F contrast gives bi-directional effects (inc/decr) so I'm
> still faced with the problem of creating one-directional contrasts
> afterward.
>
> Can you suggest anything?
> Many thanks in advance,
> Gene
Gene -
One approach is to do a conjunction or inclusive mask of the T-contrasts
[0 -1 1 0] and [0 0 -1 1]. This will find monotonic increases, though
not necessarily linear.
The more elegant approach, if you really want linearity, is to
re-specify a design matrix of one column in which condition B is coded
as -1, C is coded as 0 and D is coded as +1. Then a [+1] contrast will
find positive linear components. (If you really want to be sure of
linearity, you also have to rule out higher-order polynomial effects,
which you could do by adding a second column coded as +1, -2, +1 across
B-D, and show that a [0 1] contrast shows nothing (subject to type II
error)).
Rik
--
-------------------------------------------------------
Dr Richard Henson
MRC Cognition & Brain Sciences Unit
15 Chaucer Road
Cambridge
CB2 7EF, UK
Office: +44 (0)1223 355 294 x522
Mob: +44 (0)794 1377 345
Fax: +44 (0)1223 359 062
http://www.mrc-cbu.cam.ac.uk/people/rik.henson/personal
-------------------------------------------------------
|