Multiple Regression


I am doing regression for dependent variable Y and independent variables X1, X2, X3 and X4. Y is the weekly stock price of a company which does gold mining, processing and selling business. This company is listed in Shanghai Stock Exchange. X1 ~ X3 are the gold futures prices traded in New York, in Shanghai and US$ index. X4 is the broad market index. The correlation coefficient between X1,x2,x3,x4 with Y respectively is about 50%~60%. The number of observations are 28. The confidence level is 95%. I regressed the 5 variables, the R-square is about 47%, which is understandable, but each of the four independent variables (X1~X4) are statistically insignificant (p-value > 0.05). I don't understand. Can anyone here explain?

Thanks so much for help!



Dark Knight
It can happen.
But i think here the issue is
The number of observations are 28.
It is too small. That is the reason.
You didn't mention the model significance (F-Test). If the model is statistically significant while the predictors are not its almost always a collinearity between predictors. Also don't forget your residual assumptions.

And 28 is not neccesarilly to small.
Correlation Coefficient

One more question, if we want to use Pearson's correlation coefficient to get an idea about how two sets of data move together, the assumption is each of the two data sets is normally distributed. I have two data sets. One is with excess kurtosis = -0.68 and skewness = 0.49 while another is with excess kurtosis = -0.60 and skewness = 0.06. May I treat them as normal distribution and then use Pearson's correlation coefficient? Is there a rule of thumb for judging whether a distribution is a normal distribution or not?

Thank you guys!