1
$\begingroup$

I have a simple linear regression with age as independent variable and a cognitive scale as dependent variable. Each subject is present only once.

As it is not time-series data and there is no spatial effect, is it correct not to check for autocorrelation? Does a Durbin-Watson result of .23 mean something?

Improve this question
$\endgroup$

2 Answers 2

Reset to default
1
$\begingroup$

In general with cross-sectional data random sampling guarantees that different error terms are mutually independent, and autocorrelation is not an issue. However, when the data are collected at different hierarchical level, e.g. students within schools, or patients within hospitals, the error terms within higher-level groups may be correlated.

I'd guess that the cognitive scale depends on some factors that could be viewed as grouping factors, e.g. schooling.

Improve this answer
$\endgroup$
2
  • $\begingroup$ Both age and education influence this cognitive scale performance. So do you think a low Durbin-Watson in my regression may represent a problem due to correlation between age and some other factor (like education)? $\endgroup$
    – Aurora
    May 29, 2014 at 14:24
  • $\begingroup$ I think it could depend on the intraclass correlation between people who share the same level of education. These levels may be correlated with age among younger people, but uncorrelated among older people. If I'm guessing right (by chance) a multilevel model might help. $\endgroup$
    – Sergio
    May 29, 2014 at 14:35
0
$\begingroup$

Pearson and Durbin-Watson look very different, but can be in some cases easily and linearly, mapped back and forth between these two mathematical techniques for serial correlation. Statistical Ideas

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.