Spurious Regression

1. What is spurious regression?
Sayed Hossain posted the following figure below.

























Filippo Bonanno commented> When you regress two non-stationary variables and
there is a non-sense statistical relationship (like rain fall and income per capita in
some regions). For example if the explanatory power of the regression R-Sq is
exceptionally high, this may mean that you regressed non stationary-varianles.

Stratos D. Tserkezos commented> We have to go to Yule's (1927) paper and look
at one great example there.... nonsense correlations that are presented as
significant cause they are based on contemporaneous movements of
non-stationary series.... we can also see this with a Monte Carlo simulation of two
non-stationary series (yt= yt-1 +et and xt= xt-1 +ut) for a number of replications
and see the actual problem when we estimate y = a + bx

Idoko Gambit Itodo commented> Very high R square, low D-W stats, significant
parameters in a regression of variables that, theoretically, should not be correlated.

Takwana Chaminama commented> Nonsense regression - if R squared is high
than DW Stats there is a good reason to suspect that there is spurious regression
George Udny Yule (1927)

Tella Oluwatoba Ibrahim commented>A simple rule of model states that if a
researcher is faced with a complex and simple model, one has to go for the simple
model provided it can yield same valid result as the complex model.
Spurious regression
Hossain Academy Note
Univariate Models
Multivariate Models
Panel Data Model