Posted by
Rich Ulrich on
Jan 25, 2014; 11:24pm
URL: http://spssx-discussion.165.s1.nabble.com/studying-different-rates-of-deceleration-tp5724085p5724148.html
You do get your best test when you use an
appropriate model, so including a quadratic
term is certainly appropriate if these numbers
are locations.
Taking the use of a model just one step further:
If you *know* that you are measuring deceleration,
it could be proper to omit the linear term and
use a contrast with a declining quadratic for time,
like (16, 8, 4, 2, 1). You would include Group
as a nuisance parameter and test Group*QuadTime .l
--
Rich Ulrich
----------------------------------------
> Date: Sat, 25 Jan 2014 11:52:34 -0500
> From:
[hidden email]
> Subject: Re: studying different rates of deceleration
> To:
[hidden email]
>
> At 05:58 PM 1/22/2014, Zdaniuk, Bozena wrote:
>
>>What would be the best way to study different rates of deceleration?
>>For example, group A has means for five repeated measurements: 20, 12, 8, 6, 5
>>And group B has means: 24, 20, 18, 17, 16
>>Both groups show a decelerated curve but group A decelerates at
>>greater rate than group B.
>
> At 09:09 AM 1/23/2014, Maguin, Eugene wrote:
>
>>How about fitting a regression with a quadratic term plus a
>>group*quadratic term along with the usual group and linear trend terms?
>
> An excellent approach: summarize each time series as three values:
> . Mean
> . Linear rate of change
> . Quadratic rate of change.
>
> For this, you need an independent variable, time, across the repeated
> measurements. I'm guessing the measurements are equally spaced in
> time, so it's natural to use 1, 2, 3, 4 and 5 as the time values for
> the five points.
>
> I'd recommend, instead, centering the time variable; that is using
> values -2, -1, 0, 1 and 2. That will dramatically reduce the
> covariance of the linear and quadratic trend terms, and give you
> cleaner estimates.
>
> =====================
> To manage your subscription to SPSSX-L, send a message to
>
[hidden email] (not to SPSSX-L), with no body text except the
> command. To leave the list, send the command
> SIGNOFF SPSSX-L
> For a list of commands to manage subscriptions, send the command
> INFO REFCARD
=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD