UnThreaded | Threaded | Whole Thread (2) | Ignore Thread Prev Thread | Next Thread
Author: mikecupertino Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: of 254159  
Subject: Durbin Watson Statistic Date: 7/22/2008 6:28 PM
Post New | Post Reply | Reply Later | Create Poll Report this Post | Recommend it!
Recommendations: 0
Hello everyone,

I'm analyzing some church collections data. In order to see if I need to do time series analysis, I have computed a durbin watson statistic for the data (1.8). I need some help on how to interpret this statistic and exactly what it means if I have positive autocorrelation. Is anyone out there up on this subject?

Mikecupertino
Print the post Back To Top
Author: jj307 One star, 50 posts Old School Fool Add to my Favorite Fools Ignore this person (you won't see their posts anymore) Number: 211404 of 254159
Subject: Re: Durbin Watson Statistic Date: 7/22/2008 7:40 PM
Post New | Post Reply | Reply Later | Create Poll Report this Post | Recommend it!
Recommendations: 0
Hi Mike,
I don't think anyone else here will be interested but autocorrelation is just the correlation between two consecutive points in time in a series of one variable (like correlation is the correlation between 2 different variables at the same time).

DW is a measure of lag 1 positive autocorrelation and a value<2 should be significant.

When you do a regression the error terms are supposed to be independent. In a time series with autocorrelation this assumption is violated. If you are doing this for something like a thesis or for a large project at your job I would have somebody who knows how to do forecasting work on it.

If you are doing a short range prediction for not a ton of money and you aren't worried about things like seasonality, risk management, or CIs, I probably would not worry about the autocorrelation. -Jeff

Print the post Back To Top
UnThreaded | Threaded | Whole Thread (2) | Ignore Thread Prev Thread | Next Thread
Advertisement