Re: st: Time series panel data

["Ariel Linden, DrPH" <ariel.linden@gmail.com>]

Barth,

You haven't stated what it is you are trying to get as an end result? Are
you trying to forecast? Are you trying to find which covariates impact the
brainwaves overall or at a given point in time? Are you trying to evaluate
the effectiveness of some intervention that you begin at some point during
the night?

Without knowing what you are trying to achieve, it is impossible to give you
any meaningful advice

Ariel

Date: Wed, 28 Nov 2012 19:31:26 +0000 (UTC)
From: BARTHRILEY@comcast.net
Subject: st: Time series panel data

Dear list

I am working with a dataset consisting of sleep brainwave recordings taken
over three nights, with recordings made every 15 seconds, using a sample of
21 research participants. I have both time varying and time invariant
covariates. With this and the very long time series I am not sure of the
best model to use. Dynamic panel models such as Arellano-Bond typically
require many panels with relatively few time points, and do not take into
account the autoregressive disturbances in the data. An ARIMA model
containing only time varying predictors could be fit for each subject, and
comparisons/correlations of the intercepts and slpes generated from each
ARIMA model could be performed using the subject-level (time invariant)
variables. Is this a reasonable approach? Or is there a better alternative?

Thanks

Barth

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/