time series and comparison across groups

[Tanaka Akiko]

Hello,

I am wondering if there is a way to predict trends overtime, as well as examine variables that may predict differences in trends among different groups?

I have a task to predict number of contacts by different groups of people based on their past contacts data, but we also want to examine how various characteristics of these groups can predict the pattern (trends) or numbers.  Then we can use that knowledge of how these variables affect the number of calls to predict the number of contacts for even new groups that join (without historical data).

We have data on the variables that may affect the number of contacts or the trend, and historical data on the number of contacts at daily level, both at the group level.

The prediction of future values based on the historical data for existing groups sounds like more of time series, but using predictor variables related to differences in groups and using it to predict new groups seems more like regression to me.  Does the time series method allow us to examine differences (and what predicts these differences) across multiple groups?  I am new to time series, and I am not finding information that indicates that it can, but I thought before giving up using this approach, I would seek this group $B!G (Bs input.

Thank you.
_________________________________________________________________
5GB $B$N%*%s%i%$%s%U%!%$%kJ]B8%5!<%S%9!# (BHotmail $B$G%U%!%$%k6&M-$b$G$-$k!* (B
http://clk.atdmt.com/GBL/go/msnjpqjl0090000077gbl/direct/01/

=====================
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

[Garry Gelade]

Hello TA

It sounds like you want a growth model.  The SPSS Linear Mixed Models
procedure can handle this nicely. You can have Time as one predictor, and
analyse how the initial status and rate of change depends on your other
predictors (ie on the group characteristics).  Singer & Willett, Applied
Longitudinal Data Analysis is a good text for this type of analysis.

Garry Gelade



-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Tanaka Akiko
Sent: 04 September 2008 14:54
To: [hidden email]
Subject: time series and comparison across groups

Hello,

I am wondering if there is a way to predict trends overtime, as well as
examine variables that may predict differences in trends among different
groups?

I have a task to predict number of contacts by different groups of people
based on their past contacts data, but we also want to examine how various
characteristics of these groups can predict the pattern (trends) or numbers.
Then we can use that knowledge of how these variables affect the number of
calls to predict the number of contacts for even new groups that join
(without historical data).

We have data on the variables that may affect the number of contacts or the
trend, and historical data on the number of contacts at daily level, both at
the group level.

The prediction of future values based on the historical data for existing
groups sounds like more of time series, but using predictor variables
related to differences in groups and using it to predict new groups seems
more like regression to me.  Does the time series method allow us to examine
differences (and what predicts these differences) across multiple groups?  I
am new to time series, and I am not finding information that indicates that
it can, but I thought before giving up using this approach, I would seek
this group $B!G (Bs input.

Thank you.
_________________________________________________________________
5GB $B$N%*%s%i%$%s%U%!%$%kJ]B8%5!<%S%9!# (BHotmail
$B$G%U%!%$%k6&M-$b$G$-$k!* (B
http://clk.atdmt.com/GBL/go/msnjpqjl0090000077gbl/direct/01/

=====================
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 __________
NOD32 3414 (20080904) Information __________

This message was checked by NOD32 antivirus system.
http://www.eset.com

=====================
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