Mary,
By 'test-retest', I assume you mean the correlation between the same
measure/variable at two points in time. So, it's just a correlation.
Everything you know about correlations applies here. You need just two cases
to compute a correlation. The more cases the less a new case will shift the
value of the correlation. Given a sample of size N, a correlation greater
than or equal to a certain magnitude will be significant. Same for power.
You may have a specific hypothesis in mind: the test-rest correlation will
be at least .xx. Your sample should be large enough to give you the desired
level of confidence that you will find a correlation of that magnitude given
that it is the population value.
Gene Maguin
-----Original Message-----
From: SPSSX(r) Discussion [mailto:
[hidden email]] On Behalf Of
Martinasek, Mary
Sent: Thursday, January 06, 2011 6:08 PM
To:
[hidden email]
Subject: Help with test-retest
Can anyone provide me with references for how many cases are best for test
retest reliability and how best to analyze Likert response options?
Thanks, Mary
Mary P. Martinasek MPH, RRT
=====================
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
Locked
