Skip to main content

Table 2 Type I Error Probabilities of Methods for Missing Data Percentages (10% and 40%) and Mechanisms

From: A comparison of missing data methods for hypothesis tests of the treatment effect in substance abuse clinical trials: a Monte-Carlo simulation study

Method

Missing Percentage

MCAR

MAR

MAR/MNAR

MNAR

SSSa

10%

0.0920

0.0895

0.0875

0.0895

 

40%

0.0945

0.1035

0.1050

0.1115

Modified SSSb (Dawson Weight)

10%

0.0615

0.0580

0.0580

0.0585

 

40%

0.0525

0.0520

0.0495

0.0555

Modified SSSb (Population Weight)

10%

0.0600

0.0585

0.0585

0.0585

 

40%

0.0535

0.0520

0.0535

0.0540

Modified SSSb (SSTime Weight)

10%

0.0615

0.0590

0.0575

0.0590

 

40%

0.0540

0.0535

0.0575

0.0525

Fisherc

10%

0.0495

0.0555

0.0495

0.0510

 

40%

0.0495

0.0445

0.0505

0.0460

Weighted Z-Transformd

10%

0.0575

0.0595

0.0580

0.0570

 

40%

0.0545

0.0550

0.0525

0.0530

  1. *Type I error, r = .6, n = 100, simulations = 2000
  2. *a, b, c, d indicate the methods of analysis in the text