greg wrote:
> On Mar 14, 1:28 pm, Bruce Weaver <bwea...@[EMAIL PROTECTED]
> wrote:
>> On Mar 14, 4:30 am, greg <jsg...@[EMAIL PROTECTED]
> wrote:
>>
>>> Thanks Bruce,
>>> To be honest, I'm hoping for a linear trend in this study. A quadratic
>>> one wouldn't make much sense in this case, and with only 3 time points
>>> wouldn't be very convincing either. But I do have another study (not
>>> repeated measures, 5 ordinal groups and one continuous variable) where
>>> it looks like there might be quadratic trend (confession here to
>>> peeking at the data halfway through). It might be very useful for that
>>> one, thanks!
>> I just re-read your first post. I guess "expecting" was the wrong
>> word. It sounds like you *observed* a quadratic trend (result went
>> up, then back down).
>>
>> --
>> Bruce Weaver
>> bwea...@[EMAIL PROTECTED]
>> "When all else fails, RTFM."
>
> It looks like there may be a slight, weakly 'significant' unexpected
> up-down pattern in one variable. In an ideal world, I'd like a method
> that would take into account the fact that the groups are ordinal (in
> terms of time sequence) and I'm sure there are better methods out
> there.
Trend analysis *does* take the ordinal nature of the data into account.
(The standard form of trend analysis treats the explanatory variable
as interval scaled with a constant step size between adjacent
categories.) It is not at all appropriate in the case of a purely
nominal explanatory variable.
> I'd be happy to say that there was a quadratic relation****p if I was
> convinced there really was one, but I was concerned that by treating
> the time points as different groups, the ANOVA may be giving me a
> false positive.
--
Bruce Weaver
bweaver@[EMAIL PROTECTED]
"When all else fails, RTFM."
|
