Skip to main content

Table 5 Values of the absolute difference between the estimated and the known number of clusters across the correlation matrices for 24 variables and 96 objects

From: Thresher: determining the number of clusters while removing outliers

Methods

NbClust Top 10 Best Indices

 

Thresher

 
 

tracew

mcclain

ratkowsky

trcovw

scott

silhouette

sdindex

kl

tau

ptbiserial

CPT

TwiceMean

SCOD

1

1.008

1.011

1.055

1.078

1.213

1.528

2.246

1.850

2.382

2.320

1.147

1.153

1.011

2

1.011

1.024

1.064

1.076

1.195

2.665

2.250

1.885

2.320

2.348

0.009

0.009

1.021

3

1.018

1.013

1.070

1.088

1.221

1.674

2.285

1.829

2.369

2.344

0.037

0.037

0.988

4

0.987

0.988

0.923

0.849

0.840

1.024

0.435

0.871

0.571

0.711

2.834

2.999

1.013

5

0.982

0.994

0.936

0.908

0.845

0.962

0.424

0.947

0.550

0.695

0.790

0.307

1.008

6

0.989

0.979

0.933

0.820

0.844

1.083

0.432

0.901

0.572

0.703

2.935

2.971

1.011

7

0.011

0.012

0.067

0.137

0.206

0.614

1.254

0.838

1.369

1.360

0.760

2.634

0.318

8

0.015

0.008

0.067

0.138

0.212

0.585

1.314

0.882

1.422

1.366

0.650

2.689

0.346

9

1.014

1.016

1.050

1.133

1.198

1.562

2.250

1.902

2.365

2.320

1.163

1.163

0.981

10

1.011

1.015

1.078

1.083

1.215

1.547

2.277

1.893

2.378

2.330

1.035

1.037

1.009

11

0.023

0.020

0.094

0.741

0.231

0.571

1.226

0.981

1.373

1.308

0.050

0.049

0.350

12

0.016

0.017

0.068

0.232

0.205

0.503

1.244

0.816

1.359

1.285

0.025

0.025

0.372

13

1.983

1.985

1.920

1.597

1.779

1.634

0.875

1.497

0.839

0.951

0.878

0.881

2.006

14

1.987

1.988

1.936

1.834

1.809

1.663

0.866

1.515

0.819

0.945

1.152

1.043

2.006

15

0.983

0.988

0.928

0.805

0.831

0.942

0.410

0.861

0.554

0.699

1.853

1.858

1.033

16

0.987

0.987

0.922

0.878

0.824

0.892

0.406

0.910

0.529

0.693

1.444

1.772

1.005

Average

0.877

0.878

0.882

0.900

0.917

1.216

1.262

1.274

1.361

1.399

1.048

1.289

0.967

  1. Bold values indicate the best results for row settings