A Backward Elimination General Significance Regression Model.
ARMY MISSILE COMMAND REDSTONE ARSENAL AL SYSTEMS ANALYSIS DIV
Pagination or Media Count:
The report develops a statistical procedure for selecting a proper set of independent variables in a linear regression model. The procedure is a backward elimination procedure in which initially a large model is hypothesized and systematically non-significant variables are eliminated one by one. Two different situations were investigated concerning sample estimates of the error pure error, and using lack-of-fit as an error estimate, with the effects on the testing procedure for each case. Author
- Statistics and Probability