From: A survey of variable selection methods in two Chinese epidemiology journals
Category | Frequency | Percent |
---|---|---|
A: Select via bivariate associations | 18 | 6 |
B: Select via multivariable model | 74 | 26 |
   †Stepwise | 34 | 46 |
   †Significance testing of individual regression coefficients | 33 | 45 |
   †Unspecified | 7 | 9 |
C: Screen by bivariate associations then select via multivariable model | 85 | 30 |
   †Stepwise | 38 | 45 |
   †Significance testing of individual regression coefficients | 28 | 33 |
   †Unspecified | 19 | 22 |
D: Select via bivariate associations or multivariable model | 60 | 21 |
   †Stepwise | 25 | 42 |
   †Significance testing of individual regression coefficients | 28 | 47 |
   †Unspecified | 7 | 11 |
E: Select using other criteria | 50 | 17 |
   Prior knowledge or personal judgment | 45 | 90 |
   Tree model | 4 | 5 |
   Factor analysis followed by stepwise regression | 1 | 2 |