quantative methods
  correlation&regression
    correlation +test the significance of correlation
    linear regression line model -->regression coefficients-->SEE-->Correlation of determination  
    confidence interval -->ANOVA table -->prediction interval
  multiple regression
  model-->assumptions--->3 violations --->model misspecification

  time series analysis
  the route map :  
plot the date to see if it's linear or exponential trend
-->use DW to test to see if the serial correlation exists   
-->if not .then a trend model would be okey
-->if yes , then we choose to use AR model

before we use AR model , we need to make time series covariance stationary by doing the following
scaled the date -->take the natural log -->take the first difference

use AR model
test serial correlation with "autocorrelation of residuals "method
-->if yes ,add lag
-->if not ,test seasonality
    -->if yes ,add seasonal lag
    -->if not , test arch
             if the slope on the lagged squared residual is significant ,then ARCH exists,then we can use OLS to corret
             ARCH makes us predict the next period s squared residual               


continue tomorrow .....