I take the liberty of posting this to the Edstat list-serve for two
reasons: (1) for correction should I be offering bad advice on some
point; (2) because I do not really know SPSS well enough to answer the
specific question you are asking, and perhaps another reader of Edstat
could be more helpful in this regard. (My copy of the SPSS users' guide
is several versions out of date, and in any case it has not surfaced
since my recent move from a house to a small apartment.)
On Thu, 27 Feb 2003, Judith Bryans wrote:
> Thank you very much for replying to my query and sorry about the
> confusion.
>
> I've got 3 days and 7 time points per day. I've normalised my data
What exactly do you mean by that? The phrase is sometimes used to
indicate that one has forced a "normal" (aka Gaussian) distribution on
one or more variables, and it sometimes is used to mean other things.
Most of the "things" I know about in these contexts I wouldn't usually
recommend.
> and determined the incremental areas under the curve at each time
> point.
Areas under what curve? By "incremental" I suppose you mean "since the
previous time point", which begs the question of what you're using for a
"zero" point for your first measurement.
> I've done GLM repeated measures ANOVA and found a significant
> treatment-time effect.
Do I understand correctly that the response (aka "dependent") variable
in this analysis was those incremental areas? (I'm still having some
difficulty understanding what you did and what you think it means (or
meant), since I don't have a clear concept of what those areas are or
represent.)
> I can see from my data means and graphs that the greatest
> differences are at the latter stages of the AUC so I want to look at
> each time point with contrast and post hoc analysis.
This sounds as though the graphs of "response variable" vs. "time"
diverge from each other as time increases -- sort of a fan shape?
(At any rate, this shape would produce effects that would answer to your
description.) If this is correct, I'd be curious to know whether an
analysis of the logarithm of the response variable might produce
parallel lines. This might even be reasonable to expect, for *some*
variables; I don't know enough about "folate concentration" to guess
whether it would be reasonable for IT to behave so.
> However I'm somewhat confused because I've been advised to use the
> following syntax when doing the ANOVAs: within subject factors
> treatment add 3, time add 5, define.
I'm sorry; this makes no sense to me in the absence of a SPSS
reference. I can tell you how to carry out post hoc analyses using the
results you have so far and a calculator, but I can't tell you how to
persuade SPSS to do it for you.
> This allows me to do an ANOVA but not to do contrasts or post hoc
> analysis because I haven't defined anything as a between subject
> variable within the model if that makes sense.
Well, of course you haven't: both of your design factors (type of diet,
aka "treatments" a, b, and c; and time) are repeated measures, and
therefore within-Ss variables.
> I'm concerned that I may not have the data entered in the sheets in
> the correct manner to do the additional analysis but after SPSS
> digging through manuals and help files I'm none the wiser. At the
> moment my data is set in the following manner: I have a series of
> columns each one of which represents an incremental area at a
> particular time point for a particular treatment as follows :
> c15mins, c30mins, c45mins, c60mins, c90mins, c120mins and c150mins,
> a15mins etc, b15mins etc where c = control, a = treatment 1 and b =
> treatment 2.
I'd be curious to know how you obtained the "incremental area". Were
you using a planimeter on a graph of some kind, or did you have a
standard function that returned values (areas from "zero", perhaps)
whose differences are the "incremental areas"? Among other things, I'm
worried about how much of the effect(s) you observe are either due to
the "normalising" procedure, or partly suppressed by that procedure.
You also haven't said anything about counterbalancing (or at least
varying) the order of the three dietary treatments. If all 16 Ss
consumed the same diets in the same order (and the same number of days
apart?), there may be serious questions about the interpretations you
would like to make.
> Any advice you could give would be much appreciated.
>
> Judith
<snip, my reply to your original message...>
-----------------------------------------------------------------------
Donald F. Burrill [EMAIL PROTECTED]
56 Sebbins Pond Drive, Bedford, NH 03110 (603) 626-0816
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
