Skip to main content

Table 1 Effects of sample size on QDF for correlated 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

0.0530

0.04276

0.1240

180

0.0536

0.04656

0.1151

300

0.0543

0.04853

0.1120

450

0.0547

0.04903

0.1116

750

0.0534

0.04908

0.1089

900

0.0526

0.04857

0.1082

1200

0.0524

0.04868

0.1077

1500

0.0529

0.04893

0.1081

1800

0.0526

0.04881

0.1077

2100

0.0531

0.04913

0.1080

6000

0.0524

0.04898

0.1069

\(\delta = 2\)

90

0.0447

0.05084

0.0880

180

0.0402

0.05215

0.0771

300

0.0384

0.05224

0.0735

450

0.0394

0.05439

0.0724

750

0.0370

0.05172

0.0715

900

0.0370

0.05217

0.0710

1200

0.0369

0.05249

0.0703

1500

0.0367

0.05191

0.0707

1800

0.0372

0.05265

0.0706

2100

0.0367

0.05222

0.0702

6000

0.0365

0.05234

0.0698

\(\delta = 3\)

90

0.0300

0.6047

0.0479

180

0.0264

0.5983

0.0442

300

0.0245

0.5940

0.0412

450

0.0239

0.5907

0.0495

750

0.0232

0.5780

0.0401

900

0.0230

0.5763

0.0400

1200

0.0226

0.5708

0.0396

1500

0.0225

0.5727

0.0393

1800

0.0223

0.5682

0.0393

2100

0.0226

0.5728

0.0395

6000

0.0224

0.5698

0.0393

\(\delta = 4\)

90

0.0200

0.7509

0.0266

180

0.0159

0.7195

0.0221

300

0.0144

0.6727

0.0215

450

0.0139

0.6732

0.0207

750

0.0135

0.6574

0.0205

900

0.0131

0.6455

0.0203

1200

0.0124

0.6286

0.0198

1500

0.0126

0.6353

0.0199

1800

0.0128

0.6301

0.0203

2100

0.0128

0.6382

0.0201

6000

0.0125

0.6284

0.0200

\(\delta = 5\)

90

0.0123

0.9798

0.0126

180

0.0094

0.8717

0.0108

300

0.0084

0.8252

0.0102

450

0.0078

0.7838

0.0100

750

0.0071

0.7489

0.0095

900

0.0069

0.7399

0.0093

1200

0.0070

0.7158

0.0098

1500

0.0066

0.7058

0.0093

1800

0.0066

0.7016

0.0093

2100

0.0065

0.7002

0.0093

6000

0.0064

0.6843

0.0094