Skip to main content

Table 5 Simulation results from probability estimation for Y, where Y is generated using a probit link function with five covariates, δ is generated using a probit link function with not extreme missingness probabilities (M1) based on five covariates, N = 400

From: A nonparametric multiple imputation approach for missing categorical data

 

Pr(Y=1)=0.297

Pr(Y=2)=0.250

Method

Est

SD

SE

CR

Est

SD

SE

CR

FO

0.298

0.023

0.023

0.952

0.249

0.021

0.022

0.974

CC

0.328

0.032

0.033

0.862

0.323

0.033

0.033

0.406

 

Working models for Y:

Five covariates with logit link

  

(misspecified scenario 5)

 

Working models for δ:

Five covariates with logit link

  

(misspecified scenario 5)

CE

0.295

0.068

0.058

0.956

0.218

0.049

0.051

0.926

PMI

0.316

0.038

0.038

0.912

0.294

0.038

0.038

0.800

NNMI MLR (5,0.4,0.4;0.2)

0.310

0.039

0.040

0.940

0.275

0.036

0.039

0.930

NNMI MLR (5,0.1,0.7;0.2)

0.314

0.041

0.041

0.934

0.274

0.037

0.038

0.924

NNMI MLR (5,0.7,0.1;0.2)

0.309

0.040

0.040

0.924

0.276

0.038

0.038

0.914

NNMI CLR (5,0.4,0.4;0.2)

0.308

0.040

0.040

0.936

0.279

0.037

0.038

0.924

NNMI CLR (5,0.1,0.7;0.2)

0.305

0.039

0.040

0.930

0.279

0.037

0.039

0.914

NNMI CLR (5,0.7,0.1;0.2)

0.310

0.040

0.040

0.920

0.276

0.037

0.038

0.924

  1. Est: Estimates of probabilities; SD: Empirical standard deviation; SE: Estimate of standard error; CR: Coverage rate of 95% confidence intervals; FO: fully observed; CC: Complete Cases; CE: Calibration estimator; PMI: Parametric Multiple Imputation; NNMI MLR (NN,ω 1,ω 2;ω 3): the NNMI method using Multinomial Logistic Regressions, NN is the number of nearest neighbors and weights are ω 1,ω 2, and ω 3; NNMI CLR : the NNMI method using Cumulative Logistic Regressions; K = 10 imputed datasets are used for PMI and NNMI methods