STA 242/ENV 255: March 4, 1998

Multicollinearity

Assignment: Due Tuesday, March 10

1. CH 4: Problems 4-6
2. CH 5: Problems 8-9,
3. Do exercises 5-9 in Ch 5 for the variable "c".

1. Correlations between variables are available under Multivariate Y's. Select all variables and click on "Y", then o.k. The output will contain the table of correlations.
2. Correlations measure the strength of linear associations. Correlations are effected by outliers and may not be appropriate if the relationships are nonlinear. These will show up in scatterplots. To get all pair-wise scatter plots, go to Analyze: Scatter Plot (X Y). Select the variables of interest and add them under both the X and Y columns (in the same order). You will then get a table of pairs of plots of each Y versus each X.
3. Correlations of the parameter estimates are available by going to the Table menu after Fitting the regression model and selecting "Estimated Corr Matrix"
4. Tolerances are printed in the Box with Parameter Estimates, just after the p-values.
5. You can create the centered latitude variable within SAS INSIGHT, but it is easier to do this in the Program Editor. Find the mean of Latitude. Make sure you are not in SAS INSIGHT and then enter the following commands in the program editor: (substitue the actual numbers for "mean" below.

   data sasuser.rwg176;
   set sasuser.rwg176;
   lat2 = lat**2;
   cenlat = lat - mean;
   cenlat2 = cenlat**2;
   run;
   

   Now start up SAS INSIGHT again and proceed.
6. Durban Watson-Statistic. Calculate this by hand using the expression on page 120. To check your work, you can use the SAS PROC REG procedure.

   proc reg data sasuser.rwg137;
   model production = area year / dw;
   run;
   

   You make need to change the variables names to correpsond to how you entered them, but the Y variable goes on the left hand side of the equation and the X variables (year and area) go on the right hand side. Options for teh regression such as confidence intervals of the mean or prediction intervals (cli clm) go after the "/". In this case the "dw" tells SAS to compute the Durbin Watson Statistic.