a regression output might provide indication on whether variables are positively or negatively related. however, there is also a third possibility of neither of the aforesaid relationship are witnessed. what is the decision criteria to claim no relationship?
when the correlation (r) is zero, you can say that there is no relationship between variables. This is rare in research.
the two variables may have no relationship if:
- ESS/TSS( Estimated sum of square / Total sum of square)=r2=0
- the correlation between the two variables is zero
- there is spurious or nonsensical relationship
When a variable does not Granger cause the other variable. This resolves all spurious correlations that may arise during an investigation.
Quantitatively, It is rare to have no form of relationship at all between numerical variables i.e having exactly zero output base on their regression. The only common understanding is the variation in the degree of relationship; strong or weak
perhaps there are some problems such as serial correlation or non normality or others .
check causality between variables and test of stationarity and correlation between variables with more one lag or one lead.
hope there is useful
Dear Tesfaye
I totally agree with the above answers. But in addition, if two variables are indicated by theory that they are not correlated so that you dont have to test the correlation or test the hypothesis any more due to the spurious regression
Best
Phuc Canh
The question is: how can we solve the problem if there is no correlation no causality no effect between variables?
- Badges
- Science topic