Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: Time series analysis


From   David Greenberg <dg4@nyu.edu>
To   statalist@hsphsun2.harvard.edu, aharri@calgary.edu
Subject   Re: st: Time series analysis
Date   Mon, 15 Oct 2007 13:02:55 -0400

With just five subjects, but many time points, you have far fewer cases than is typical in a panel data set. Technically, you could use panel-corrected standard errors (pcse is the command in Stata), but I would be tempted to exploit the fact that youhave so many time points and so many cases by doing separate time series analyses for each case. That way, with ARIMA modeling, you could learn much more about the autoregressive and moving average components of your data than yo would with a typical panel model approach. Then you could compare the results you get for the five subjects. Of course, if you take this suggestion, you will need to impute values for the missing observations. David Greenberg, Sociology Department, New York University

----- Original Message -----
From: Ashley Harris <aharri@ucalgary.ca>
Date: Saturday, October 13, 2007 5:47 pm
Subject: st: Time series analysis
To: statalist@hsphsun2.harvard.edu


> Hello Statalist, 
> 
> I have done a lot of reading on time-series with interventions, but 
> have a
> hard time interpreting the different situations and the specific translation
> to Stata-language, so am looking for advice, guidance etc.
> 
> 
> I have data on 5 subjects with data over time - every 5-15 mins I have 
> data
> point. There are a variety of parameters that I am looking at, hence the
> data points. 
> 
> I cause an intervention on one side - ie. Right or left, and then the 
> other
> side is my control- arbitrarily I have set my intervention to time = 
> 0. The
> nature of my intervention causes a transient response.
> 
> So, what I want to do is first see if I actually see a response to the
> intervention. So, I start with:
> 
> tsset sub_num time
>        panel variable:  sub_num, 1 to 5
>         time variable:  time, -5 to 150, but with gaps
> 
> And then 
> Xtreg right_result intervention, re
> 
> Where right_result I the data for the right side and intervention is 0 
> if
> time < 0 and intervention = 1 for time>0
> 
> 
> 
> Now, here come the questions:
> 
> It seems to me xtreg is the most straight forward time-series analysis 
> for
> panel data or am I simplifying something that I shouldn't be? - how do 
> I
> know? 
> 
> For my results from a parameter that is not normal - is there an alternative
> I can use?
> 
> Lastly - if I want to find out when the maximum response to my intervention
> occurs and when that response goes away, what do I do?
> 
> Thank you in advance!!!
> Ashley
> 
> 
> *
> *   For searches and help try:
> *   http://www.stata.com/support/faqs/res/findit.html
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index