[R] simple repeated measures model: dumb user baffled!

Chris Evans stats at psyctc.org
Wed Apr 28 22:29:09 CEST 2004

I am in the process of transferring from an old version of S+ to using
R having used a variety of other packages in the past.  I'm hugely
impressed with R but it has an excellent but depressing habit of exposing
that I'm not a professional statistician and has done so again.

Someone has run a nice little repeated measures design on my advice,
students randomised to four orders of a condition that can be N or E:
four orders used: NEEN, NENE, ENEN, ENNE, ten students in each block.  I've inherited
the data (in SPSS but I can deal with that!) with variables like:

ORDER is the order as a/b/c/d; C1:C4 has the N or E for each occasion, and
RESP1:RESP4 the response variables.  (There are a number of these but
looking at each separately is justifiable theoretically).

I've had a look around the R help and some S+ books I've got and I
realise I'm seriously out of my depth and my repeated measures ANOVA
knowledge is rusty and very different from the way that more modern
statistics handles such designs.

Can anyone point me to an idiot's guide to the syntax that would help
me test:
a) that there is a change (probably a fall in RESPn) over the four
repeats (probable through a practice effect)
b) whether that shows any sign of higher than linear change
c) whether on top of that, there are N/E differences.

I realise that this is probably trivially easy but I'm staring at all
sorts of wonderful things in Venables & Ripley (S+ 2nd ed.) and in Chambers & Hastie
(S, 1st ed.) but nothing is quite near enough to what I need to help
me overcome my limitations!



More information about the R-help mailing list