Remedial measures:
There are two common approaches taken towards the problem of multicollinearity. First, 'do nothing' which is simply to ignore the problem. Alternatively, the researcher can follow some rules of thumb as outlined below. The order in which the measures are given here is roughly by descending order,of their popularity. For instance, most researchers choose the 'do nothing' approach if there are no serious problenk in interpreting the coefficients, and so on.
Do nothing : Since multicollinearity is a problem with the data at hand, it is not possible to do anything about it, and therefore we try to make the most of existing situation. In other words, it is a sample problem; the sample available does not contain enough variation to, identify certain impacts (for instance, if wealth and income always vary proportionately then the impact of both separately on consumption cannot be identified). Usually not all coefficients are statistically insignificant in a regression model. Hence, we work with what we can get out of the model and hope that availablity of more data will allow a more accurate estimation in future. Moreover, even if one or more regression coefficients cannot be estimated with great precision, a linear combination of them can be estimated relatively efficiently. For instance, from equation above if we know that X3=W2, then substituting this into the equation we get
where. Note that even if we knew λ. in the case of perfect collinearity, there is no way to get the two cofficients β, and β, separately. since we have one equation with t vo unknowns . Sometimes we have to accept that this is the best that we can do given the data.