SPURIOUS CORRELATION AND IT'S DETECTION
Consider a case where the X variable causes Y one period
in the future. Consider a country where the Income in year "Z"
effects or is useful in predicting the consumption one year in
the future. X is referred to as a lag variable. The
question is how can this lag be detected and subsequently harnessed
in order to better describe the behavior of Y ?
Initially, if we run a simple multiple regression, we get the
following results.
| |
If we examine the autocorrelation of the residuals we
find that they do not appear to have any structure. Most analysts
would stop here.
| |
However for a causal model to be sufficient one needs to
check as to whether or not any remaining or auxiliary information
exists in lags of the series of interest. We find a relatively
high value of .65 indicating a need to change the model.
| |
Estimating the new model where the additional lag
is found to be significant we get...
| |
Finally the proof of sufficiency ... both the ACF and the
CCF are approximately zero indicating that there is no need for any
more changes.
| |
The data.
| |
CLICK HERE:Home Page For AUTOBOX