Skip to main content

Table 3 Simulation study with uniform and gamma models

From: Comparison of confidence interval methods for an intra-class correlation coefficient (ICC)

 

Uniform model

Gamma model

b0= 48, l0= 3, r0= 1

b0= 96, l0= 6, r0= 1

Low skew

High skew

ICC w

Method

Cov.

Wid.

Cov.

Wid.

Cov.

Wid.

Cov.

Wid.

0.99

GPQ

0.976

0.768

0.986

0.544

0.938

0.749

0.918

0.731

MLS

0.977

0.771

0.986

0.546

0.941

0.752

0.922

0.734

Bayes

0.873

0.823

0.985

0.593

0.856

0.825

0.849

0.823

0.90

GPQ

0.977

0.705

0.985

0.464

0.935

0.684

0.919

0.670

MLS

0.979

0.710

0.986

0.467

0.937

0.689

0.926

0.675

Bayes

0.879

0.790

0.986

0.516

0.849

0.788

0.843

0.767

0.80

GPQ

0.980

0.614

0.990

0.344

0.931

0.600

0.901

0.586

MLS

0.981

0.621

0.990

0.347

0.936

0.607

0.908

0.593

Bayes

0.883

0.753

0.992

0.394

0.932

0.752

0.805

0.745

0.71

GPQ

0.987

0.379

0.994

0.156

0.903

0.422

0.866

0.421

MLS

0.992

0.384

0.996

0.156

0.918

0.429

0.884

0.428

 

Bayes

0.958

0.686

0.996

0.169

0.859

0.699

0.832

0.695

  1. Comparison of MLS, GPQ and Bayes method performance on uniform and gamma data. Nominal 95% confidence intervals for ICCb. Coverages and average widths calculated from 10,000 simulations. In each case, ICCb = 0.70. Study designs have 48 biological replicates and 3 labs for a total of 144 observations, and 96 biological replicates and 6 labs for a total of 576 observations. Means and standard deviations of the point estimates of the ICC b for each setting are presented in a Additional file 2: Excel file.