Skip to main content
. 2021 May 27;40(20):4362–4375. doi: 10.1002/sim.9034

TABLE 2.

Estimated LASSO (least absolute shrinkage and selection operator) shrunk coefficients and coefficients from the prespecified model together with penalized maximum likelihood estimation

Variables LASSO model coefficients Prespecified model coefficients (SE)
C‐score 0.60 0.62
Calibration slope 1.54 1.05
Intercept −0.4424 −0.8656 (0.866)
Age −0.0013 −0.0181 (0.005)
Sex (male vs female) −0.1379 (0.092)
Baseline weight −0.0002
Baseline EDSS 0.0963 0.1683 (0.047)
Years since onset of symptoms 0.0587 (0.063)
Ethnicity (white vs other) −0.0142 (0.117)
No. of relapses 1 year prior to study 0.2971 0.5963 (0.170)
Months since prestudy relapse −0.0126 (0.009)
Prior MS treatment group (yes vs no) 0.0241 0.1901 (0.085)
Region (India vs Eastern Europe) 0.0000
Region (North America vs Eastern Europe) 0.0000
Region (Rest of world vs Eastern Europe) 0.0000
Region (Western Europe vs Eastern Europe) 0.2374
Timed 25‐Foot Walk −0.1718 (0.158)
9‐Hole Peg Test 0.3011 (0.208)
PASAT‐3 0.0029 (0.004)
VFT 2.5% −0.0010 (0.004)
Baseline Gadolinium volume 0.0001
Baseline SF‐36 PCS −0.0120 −0.0195 (0.005)
Baseline SF‐36 MCS 0.036 (0.004)
Baseline actual distance walked (>500 vs ≤500) −0.0746

Note: The discrimination (C‐score) and the calibration slopes are also shown.

Abbreviations: EDSS, expanded disability status scale; MS, multiple sclerosis; PASAT, paced auditory serial addition test; SE, standard error; SF‐36 MCS, short form‐36 mental component summary; SF‐36 PCS, short form‐36 physical component summary; VFT, visual function test.