Skip to main content
. 2021 Feb 22;12:611506. doi: 10.3389/fgene.2021.611506

Table 2.

Stability of feature selection methods for prediction of residual feed intake with LASSO regression.

Subset size1 Filter2 NSNPs3 PDF4 medianSel5 IQRSel6 NOG7
50 mrmr 499 0.30 50 0 0.53
cforest 454 0.89 46 5 0.03
spearcor 441 0.20 44 3 0.69
univ.dtree 151 0.99 24 10 0.00
random 189 0.99 23 12 0.00
250 mrmr 2,096 0.28 209 8 0.53
cforest 1,693 0.79 172 13 0.05
spearcor 1,575 0.28 158 12 0.54
univ.dtree 804 0.90 83 18 0.02
random 740 0.92 80 48 0.01
500 mrmr 3,420 0.29 347 23 0.52
cforest 2,287 0.72 239 39 0.07
spearcor 2,560 0.28 260 9 0.52
univ.dtree 1,292 0.86 141 40 0.02
random 1,332 0.86 151 63 0.02
750 mrmr 4,333 0.28 437 24 0.53
cforest 2,646 0.67 275 35 0.09
spearcor 3,339 0.27 338 21 0.53
univ.dtree 1,667 0.79 165 40 0.04
random 1,745 0.80 160 57 0.03
1,000 mrmr 5,125 0.27 515 23 0.54
cforest 3,008 0.63 321 53 0.10
spearcor 3,978 0.27 394 15 0.53
univ.dtree 1,750 0.74 193 21 0.06
random 1,879 0.76 192 62 0.05
1,500 mrmr 5,978 0.26 595 21 0.55
cforest 3,478 0.56 341 8 0.14
spearcor 5,096 0.27 506 12 0.54
univ.dtree 2,270 0.66 235 19 0.09
random 2,200 0.68 235 80 0.08
9,523 none 2,963 0.27 269 111 0.56
1

Subset size = number of selected features.

2

Filter method = Maximum relevance minimum redundancy (mrmr); Random forest (cforest); Spearman’s correlation (spearcor); Univariate decision tree (univ.dtree); Random selection (random).

3

NSNPs = Total number of SNPs pre-selected in the 10 subsets.

4

PDF = Proportion of distinct features in the 10 subsets.

5

medianSel = Mean number of selected SNPS in the 10 subsets.

6

IQRSel = Interquartile range of the number of selected SNPS in the 10 subsets.

7

NOG = Nogueira et al. (2018) stability estimator.