Hi Laura,
Yes with grouped data you explore patterns by group e.g., males vs
females. However, I recall that Tabachnick and Fidell don't tell you
what to do if some of your analyses involve grouped data and others
ungrouped. I chose to screen for group first then re-ran all analyses
for the whole sample. Have you had any other responses to your e-mail?
Note that in regression the assumption is normality of the residuals and
outliers can be identified as part of the analysis.
Kathryn
>>> Laura Ashley <[log in to unmask]> 15/05/2007 11:44:11 >>>
Hi All,
I am performing multiple regression with a categorical predictor and a
continuous predictor (continuous outcome variable), and data screening
following Tabachnick & Fidell.
They note that there are differences in data screening for grouped and
ungrouped
data. They say that if you are performing multiple regression you are
screening
ungrouped data.
But as I have a categorical predictor in my regression (experimental
condition),
my data are grouped (?).
So am I right to think that I should screen as for grouped data rather
than
ungrouped? (that is, look for outliers, check for normality etc. within
each
group separately rather than among all cases at once).
Any advice on this would be great.
Thanks,
Laura
|