Serial Correlation (Univariate models)

1. If there is lag in model, we can not use Durbin Watson test
Professor Noman Arshed commented> Durbin watson test is applicable for time series
autocorrelation testing but it has one limitation that if there are lags in model, we cannot
use this.

2. Disadvanatages of Durbin-Watson test
Professor Nicat Gasim commented as such> Durbin Watson test has three
disadvantages. 1. If models comtain lagged variables DW test cannot use for detecting
autocorrelation. 2.It has indesicion area. 3. DW test derermine only first order
autocorrelation


3. Normality, autocorrelation and heteroscedsaticity
Professor Sudarshan Bhattacharjee​ commented about diagnostic checking as such >> If
the data series is sufficiently large... about more than 30 0bservations then you can ignore
the violation of normality assumption coz by central limit theorem a long series will
converge to normality.... for auto-correlation and heteroskedasticity we can try first some
transformation in the variables.... however, if that does not work then use Newey-West
test for HC and HAC... This is readily available in R and Stata... Not sure about Eviews....
hope this helps..

4. How to eliminate serial correlation fro the time series model?
Profesor Muzammil Bhatti commented as such >> A special type of GLS ie. WLS is the
best option instead of lag of dependent because it just solves first order serial corelation.

5. Consequeces of serial correlation
Professor Osifo Osarumwense commented as such>> A model suffering from serial
correlation is likely to have understated or overstated standard errors of parameter
estimates. In either case, this can negatively affect our calculated t-values and hence our
test of hypothesis. In summary, it may lead us to draw false conclusion about the
statistical significance or otherwise of parameter estimates. For this reason, I cannot
accept a model with serial correlation.

6. How to eliminate serial correlation from the model?
Professor Sayed Hossain commented as such>
























7. Removal of autocorrelation
Professor Louis Atamja posted the question as below:
"I was studying how to remove auto correlation in a time series and i found out that we
can take the first difference. What if after taking the first difference there is still
autocorrelation. What should i do?"

Seye Olasehinde-Williams commented> Just run the regression with hac
(heteroscrdasticity and autocorrelation consistent) standard errors. The option is

Muzammil Bhatti commented>First difference just removes first order auto correlation.
HAC is good option while in time series you should also introduce autoregressive part, in
this way your model not only becomes a dynamic model but also ensures almost no
autocorrelation.

8. What is autocorrelation?
Olasehinde Timilehin commented> For an instance, previous government debt will surely
have impact on the current government debt. If government failed to repay back in full
(interest and principal) the debt of last year, surely the outstanding debt will be carried
over to the current year. In accounting, it is termed, balance brought down /forward.

9. What is autocorrelation?
Aadersh Joshi commented> Autocorrelation is a mathematical representation of the
degree of similarity between a given time series and a lagged version of itself over
successive time intervals. It is the same as calculating the correlation between two
different time series, except that the same time series is used twice: once in its original
form and once lagged one or more time periods.

10. How to avoid serial correlation in the model?
Asad Zaman commented>
Serial correltion in Univariate Model
Hossain Academy Note
Univariate Models
Multivariate Models
Panel Data Model