Skip to main content

Table 3 Parameter estimates for one-part model, two-part (TP) model with uncorrelated random effects, TP with correlated random effects, and marginalized two-part (mTP) models assuming Weibull distribution for the non-zero component

From: Analysis of longitudinal semicontinuous data using marginalized two-part model

Model component

Covariate

One-part modela,b: parameter estimate, (SE), P-value

TP model uncorrelated random effectsc: parameter estimate, (SE), P-value

TP model correlated random effectsd: parameter estimate, (SE), P-value

mTP modele: parameter estimate, (SE), P-value

Zero part

Intercept

3.8358, (1.9236), 0.0465

0.5006, (0.1246), < 0.0001

− 0.2263, (0.2050), 0.2700

− 0.1352, (0.1160), 0.2265

Txt group

0.2915, (0.3266), 0.3725

− 0.2506, (0.1496), 0.0944

− 0.3951, (0.2021), 0.0509

− 0.3018, (0.1205), 0.0125

Smoking

− 0.7732, (0.3887), 0.0471

0.5177, (0.1944), 0.0080

0.5437, (0.2489), 0.0293

0.4978, (0.1708), 0.0037

Time

− 0.0384, (0.0537), 0.4747

0.0355, (0.0190), 0.0626

− 0.2700, (0.0387), < 0.0001

− 0.0768, (0.0192), < 0.0001

Continuous non-zero part

Intercept

–

2.4153, (0.3739), < 0.0001

− 1.9956, (1.4045), 0.1558

− 1.9886, (0.7085), 0.0051

HbA1c

− 0.2964, (0.0905), 0.0011

− 0.0076, (0.0187), 0.6844

0.0988, (0.0672), 0.1418

0.0828, (0.0380), 0.0296

Age

− 0.0552, (0.0241), 0.0221

− 0.0121, (0.0044), 0.0071

0.0028, (0.0169), 0.8644

− 0.0017, (0.0086), 0.8367

Duration

− 0.0099, (0.0358), 0.7816

− 0.0094, (0.0069), 0.1749

0.0088, (0.0235), 0.7061

0.0003, (0.0141), 0.9832

SBP

0.0010, (0.0128), 0.4348

0.0055, (0.0025), 0.0303

0.0322, (0.0091), 0.0005

0.0231, (0.0054), < 0.0001

Male

− 0.7926, (0.3534), 0.0252

0.0453, (0.0673), 0.5010

0.0006, (0.2411), 0.9980

0.0057, (0.1352), 0.9659

HDL

0.0209, (0.0131), 0.1091

0.0058, (0.0025), 0.0230

0.0193, (0.0094), 0.0400

0.0130, (0.0053), 0.0150

Time

–

0.0235, (0.0109), 0.0314

0.0099, (0.0373), 0.7888

− 0.0282, (0.0257), 0.2716

Random effects

Zero part variance

–

1.0528

0.6612

1.3971

Non zero part variance

–

0.2094

1.6636

1.2199

Covariance

–

–

1.0488

1.3055

  1. One-part modela fits the entire sample without distinction between zero and non-zero processes, so only one estimate for the intercept and one for time were generated. In one-part modelb the parameter estimates for txt group and smoking represent the effect of these covariates on the CTGF levels themselves and not on the probability of non-zero values, unlike the TP and mTP models. TP model uncorrelated random effectsc and TP model correlated random effectsd generate estimates for the continuous part using only a portion of the sample pertaining to positive non-zero values. mTP modele provides estimates for the parameters in the continuous part for the entire sample (zero and non-zero values)
Back to article page

Journal of Translational Medicine

ISSN: 1479-5876

Contact us