Table 3 Mean over m = 500 dose + cancer simulations of regression coefficients in fits of linear-quadratic model.

From: A generalisation of the method of regression calibration

Magnitude of error distribution (GSD)

Unadjusted

Regression calibration

Extended regression calibration

Monte Carlo maximum likelihood

ERR/Gy

ERR/Gy2

ERR/Gy

ERR/Gy2

ERR/Gy

ERR/Gy2

ERR/Gy

ERR/Gy2

Unshared Berkson error (%)

Shared Berkson error (%)

Unshared classical error (%)

Shared classical error (%)

α

β

α

β

α

β

α

β

0

0

20

20

0.221

2.278

0.196

2.061

0.196

2.061

0.196

2.061

0

0

20

50

0.288

4.168

0.196

2.061

0.196

2.061

0.196

2.061

0

0

50

20

0.255

2.260

0.196

2.061

0.196

2.061

0.196

2.061

0

0

50

50

0.328

4.136

0.196

2.061

0.196

2.061

0.196

2.061

20

20

20

20

0.220

2.469

0.195

2.233

0.125

2.132

0.288

2.207

20

20

20

50

0.287

4.523

0.195

2.233

0.125

2.132

0.288

2.207

20

20

50

20

0.255

2.451

0.195

2.233

0.125

2.132

0.288

2.207

20

20

50

50

0.328

4.492

0.195

2.233

0.125

2.132

0.288

2.207

20

50

20

20

0.262

2.983

0.227

2.707

0.109

2.393

0.370

3.007

20

50

20

50

0.354

5.426

0.227

2.707

0.109

2.393

0.370

3.007

20

50

50

20

0.303

2.962

0.227

2.707

0.109

2.393

0.370

3.007

20

50

50

50

0.401

5.390

0.227

2.707

0.109

2.393

0.370

3.007

50

20

20

20

0.259

2.986

0.224

2.709

0.121

2.354

0.337

2.678

50

20

20

50

0.347

5.441

0.224

2.709

0.121

2.354

0.337

2.678

50

20

50

20

0.299

2.964

0.224

2.709

0.121

2.354

0.337

2.678

50

20

50

50

0.395

5.401

0.224

2.709

0.121

2.354

0.337

2.678

50

50

20

20

0.243

3.703

0.209

3.349

0.038

2.795

0.362

3.401

50

50

20

50

0.332

6.744

0.209

3.349

0.038

2.795

0.362

3.401

50

50

50

20

0.286

3.682

0.209

3.349

0.038

2.795

0.362

3.401

50

50

50

50

0.383

6.703

0.209

3.349

0.038

2.795

0.362

3.401

True value

   

0.25

2.0

0.25

2.0

0.25

2.0

0.25

2.0

  1. GSD geometric standard deviation, ERR excess relative risk.