From: New adaptive lasso approaches for variable selection in automated pharmacovigilance signal detection
Scenario | β0 | nTP | βTP | Average number of covariates retained after filtering | Average number of true predictors retained after filtering | Average number of cases |
---|---|---|---|---|---|---|
No true predictors | ||||||
1 | -2 | 0 | 0 | 389 | NA | 11 923 |
2 | -4 | 0 | 0 | 165 | NA | 1 798 |
3 | -6 | 0 | 0 | 30 | NA | 247 |
True predictors reported more than 100 times | ||||||
4 | -2 | 5 | 1 | 394 | 5 | 12 327 |
5 | -2 | 5 | 2 | 400 | 5 | 12 937 |
6 | -2 | 20 | 1 | 407 | 20 | 13 589 |
7 | -2 | 20 | 2 | 424 | 20 | 15 956 |
8 | -4 | 5 | 1 | 172 | 5 | 1 879 |
9 | -4 | 5 | 2 | 184 | 5 | 2 083 |
10 | -4 | 20 | 1 | 193 | 20 | 2 155 |
11 | -4 | 20 | 2 | 237 | 20 | 3 121 |
12 | -6 | 5 | 1 | 34 | 2 | 259 |
13 | -6 | 5 | 2 | 44 | 4 | 294 |
14 | -6 | 20 | 1 | 49 | 11 | 300 |
15 | -6 | 20 | 2 | 95 | 18 | 507 |
True predictors reported between 20 to 100 times | ||||||
16 | -2 | 5 | 1 | 391 | 5 | 11 958 |
17 | -2 | 5 | 2 | 392 | 5 | 12 013 |
18 | -2 | 20 | 1 | 396 | 20 | 12 064 |
19 | -2 | 20 | 2 | 400 | 20 | 12 284 |
20 | -4 | 5 | 1 | 167 | 2 | 1 805 |
21 | -4 | 5 | 2 | 171 | 4 | 1 822 |
22 | -4 | 20 | 1 | 175 | 8 | 1 827 |
23 | -4 | 20 | 2 | 191 | 17 | 1 897 |
24 | -6 | 5 | 1 | 30 | 0 | 248 |
25 | -6 | 5 | 2 | 31 | 0 | 251 |
26 | -6 | 20 | 1 | 31 | 0 | 251 |
27 | -6 | 20 | 2 | 35 | 2 | 263 |