Spline regression & autorcorrelation

[tino]

Hello everybody!

 I was wondering if anyone might give me a tip how to deal with autocorrelation when using SPSS for doing segmented regression analysis (aka spline regression) for longitudinal data?

Specifically, I have a small time series with n = 30 observations(just one observation per measurement point).

While these data which are more or less linearly related, they contain one clearly visible discontinuity.

Modeling this “jump” appears to be quite straightforward – however, I am struggling with the potential autocorrelation in the data. The Durbin-Watson test statistics is around 1.1, which I take as an indicator of positive autocorrelation (which is also apparent from the data). Is there any procedure available in SPSS which might account for autocorrelation, preferably outside the ARIMA-approach?

Many thanks for your reply,

Tino

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

[Jon K Peck]

Is the discontinuity in the errors?  While you could model it in various ways, understanding the cause might be important.

As for autocorrelation, you can use the AREG procedure to handle AR(1) errors.


Jon Peck (no "h") aka Kim
Senior Software Engineer, IBM
[hidden email]
phone: [phone]




From:        Tino Nsenene <[hidden email]>
To:        [hidden email]
Date:        08/13/2014 02:37 AM
Subject:        [SPSSX-L] Spline regression & autorcorrelation
Sent by:        "SPSSX(r) Discussion" <[hidden email]>

---

Hello everybody!

I was wondering if anyone might give me a tip how to deal with autocorrelation when using SPSS for doing segmented regression analysis (aka spline regression) for longitudinal data?

Specifically, I have a small time series with n = 30 observations(just one observation per measurement point).

While these data which are more or less linearly related, they contain one clearly visible discontinuity.

Modeling this “jump” appears to be quite straightforward – however, I am struggling with the potential autocorrelation in the data. The Durbin-Watson test statistics is around 1.1, which I take as an indicator of positive autocorrelation (which is also apparent from the data). Is there any procedure available in SPSS which might account for autocorrelation, preferably outside the ARIMA-approach?

Many thanks for your reply,

Tino

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