*Program  for Manufacturing Firm per-unit cost 
*Sara Gieseke
*November 2004
*Program name:  per-unit cost;

/*variable used in the model
cost function for 20 years
unitcost = cost per unit in dollars (range 3.65-6.62)
output = index of output (range 50-104)
inpcost = index of input costs (range 80-150)
output2 = output^2
inpcost2 = inpcost^2
*/

proc print data=cost;
run;

/*DESCRIPTIVE STATISTICS*/

/*produce basic summary statistic*/
proc means data=cost;
run;
/*produce detailed summary statistics, tests of normality, stem-and-leaf plots
and boxplots*/
proc univariate data=cost normal plots;
	title 'Summary statistics for Manufacturing Firm';
run;
/*use proc insight to create boxplots of the data*/
proc insight data=cost;
	box unitcost output output2 inpcost var6;
run;
quit;

/*CORRELATION TESTS*/

/*use proc insight to create a correlation matrix*/
proc insight data=cost;
	scatter unitcost output output2 inpcost var6 *unitcost output output2 inpcost var6;
run;
quit;

proc corr data=cost;
	title 'Correlation matrix of Manufacturing variables';
	var unitcost output output2 inpcost var6;
run;

proc insight data=cost;
	scatter unitcost output output2 inpcost var6*unitcost output output2 inpcost;
run;
quit;

/*REGRESSION*/

proc reg data=cost;
	title 'Regression of Manufacturing Unit Cost';
	model unitcost=output output2 inpcost var6 / selection=forward slentry=0.05;
/*best model is re-run below*/
/*print out of predicted and residuals*/
proc reg data=cost;
	title 'Best fit model';
	model unitcost=output output2 inpcost/ p r;
/*plot residuals to check assumptions*/
	plot residual.*predicted.='0';
run;