Skip to main content

Table 2 Effects of sample size on QDF for uncorrelated normal based on error rates, CV and SD

From: Robustness of the Quadratic Discriminant Function to correlated and uncorrelated normal training samples

Centroid

Sample size (n)

SD

CV

Mean error rate

\(\delta = 1\)

90

0.0681

0.3483

0.1955

180

0.0641

0.3920

0.1636

300

0.0603

0.3965

0.1522

450

0.0603

0.4092

0.1473

750

0.0575

0.4059

0.1417

900

0.0573

0.4065

0.1409

1200

0.1387

0.4096

0.1387

1500

0.0571

0.4114

0.1388

1800

0.0567

0.4123

0.1374

2100

0.0567

0.4129

0.1374

6000

0.0559

0.4116

0.1359

\(\delta = 2\)

90

0.0524

0.3968

0.1321

180

0.0469

0.4222

0.1111

300

0.0421

0.4062

0.1037

450

0.0424

0.4251

0.0997

750

0.0409

0.4218

0.0970

900

0.0415

0.4313

0.0962

1200

0.0402

0.4209

0.0955

1500

0.0398

0.4219

0.0944

1800

0.0400

0.4230

0.0946

2100

0.0399

0.4239

0.0940

6000

0.0394

0.4223

0.0934

\(\delta = 3\)

90

0.0345

0.4200

0.0823

180

0.0296

0.4341

0.0683

300

0.0277

0.4417

0.0028

450

0.0263

0.4395

0.0599

750

0.0256

0.4372

0.0586

900

0.0257

0.4385

0.0586

1200

0.0255

0.4393

0.0581

1500

0.0253

0.4386

0.0578

1800

0.0248

0.4330

0.0573

2100

0.0250

0.4361

0.0574

6000

0.0249

0.4380

0.0568

\(\delta = 4\)

90

0.0236

0.4866

0.0486

180

0.0189

0.4991

0.0379

300

0.0173

0.4889

0.0354

450

0.0160

0.4802

0.0334

750

0.0158

0.4815

0.0327

900

0.0153

0.4744

0.0323

1200

0.0151

0.4725

0.0319

1500

0.0151

0.4728

0.0318

1800

0.0147

0.4660

0.0315

2100

0.0147

0.4663

0.0315

6000

0.0146

0.4656

0.0313

\(\delta = 5\)

90

0.0156

0.5989

0.0260

180

0.0121

0.5892

0.0205

300

0.0098

0.5613

0.0174

450

0.0093

0.5367

0.0173

750

0.0090

0.5507

0.0163

900

0.0085

0.5262

0.0161

1200

0.0085

0.5260

0.0162

1500

0.0085

0.5272

0.0161

1800

0.0083

0.5182

0.0160

2100

0.0083

0.5221

0.0159

6000

0.0081

0.5109

0.0158