Patty,
I think your dataset must be a little more complicated than you describe
because you mention '6 variables'. If the 6 variables are days then you
would have days nested with persons. But, if the six variables are just six
variables (e.g., height, weight, etc) measured each day then you have a
three level structure: measures within days within persons. This is
something to clarify.
The intraclass correlation (ICC) is a property of the sample. Two of many
references here are McGraw and Wong (1996), Psychological methods, and
Shrout and Fleiss (1979). It seems that you are thinking of it as a property
of the person. If you'll read the McGraw and Wong article you'll see that
there are several different definitions for ICC depending on whether you are
interested in a single item or a set of items and how 'persons' are to be
treated. I think there is some documentation of this in the syntax
reference. Maybe it'd be helpful to explain the backgroud for computing the
ICC and what you are wanting the computational result to accomplish.
Gene Maguin
________________________________
From: SPSSX(r) Discussion [mailto:
[hidden email]] On Behalf Of
Patty Zorbas Ferssizidis
Sent: Tuesday, March 22, 2011 11:00 AM
To:
[hidden email]
Subject: Linking OMS output to unique ID's in primary database
I have a data set that contains multiple rows (i.e, days) of data per
participant (i.e., ID). I am trying to compute the intraclass correlation
coefficient (based on 6 variables) for each unique ID across the multiple
entries/days for this ID. Then I need to link the coefficient to each unique
ID in the primary database so that it can be used for further analysis.
Using a split file command (to separate output by ID) and the OMS feature
(with dimensions in a single row), I have been able to save a new data file
containing the coefficients. However, the coefficients are not linked to
each unique ID and therefore, cannot be merged into the primary database and
matched by ID. Because the ICC does not compute for every ID (e.g., some
ID's have no data or for some reason cannot be calculated), I get fewer
ICC's than I have ID's. Is there a way to match the ICC from the output to
the ID from which it was computed? If that is not possible, I suppose an
alternative would be to create syntax to compute the ICC's (any ideas on how
to go about that would also be helpful if it's my only option). Many thanks
in advance.
Patty
=====================
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
