1-Introduction
Several years ago some researchers in Econometric field used ordinary least square
(OLS) method to estimate the parameters of linear regression model. The Ordinary least
square method is based on a number of assumptions; one of these assumptions says that
there is no linear relationship between the explanatory variables, in case of dropping this
hypothesis the multicollinearity problem will appear. The multicollinearity problem
threatens both the assumption and usage of ordinary least square (OLS). In case the model
is not full ranked,it won’t be possible to find the inverse matrix of explanatory
variables;consequently, we will get infinite solutions according to (OLS) method. On the
other hand if the relationship degree between the explanatory variables is much enough
and we use the (OLS) method we will find that:
• The estimations of model parameters have different results in addition to that the
standard errors for these estimations increase when the relationship between
explanatory variables increases.
• More standard errors of estimators mean a breadth of confidence interval in the
parameters of pobulation.
• Increasing risk of type two errors (accepting wrong assumptions) because of the
width of confidence interval due to an increase in the standard error of the
estimation.
• If the linear relationship is high, we get higher R2 however;most parameters of the
relationship are non-significant statistics by using t test.
The existence or nonexistence of the linear relationship between the variables isn’t the
point but the strength degree is .We previously mentioned that there is a linear
relationship between the explanatory variables (mathematical variables) not a random
relationship, so this phenomenon concerns with the sample not the society for whom
the sample was selected.So we don’t test the existence of these relationships but we
measure the strength degree in any sample.Due to the negative impact caused by
Multicollinearity problem, there are many ways to deal with this problem and one of
these methods is Ridge regression (RR) .