Skip to main content

Table 4 Goodness of fit for latent class models (n = 47,004)

From: Testing extra-linearity across a psychosis continuum

Number of classes

Log Likelihood

Number of parameters

AIC

BIC

AdjBIC

LMRT P-value

Entropy

BLRT P-value

No. classes with n < 5% study sample

Level 1–5

 2

− 156,499

51

313,100

313,547

313,385

0.33

0.94

0.00

0

 3

−151,298

77

302,751

303,425

303,180

0.00

0.90

0.00

1

 4

−148,819

103

297,844

298,746

298,419

0.00

0.91

0.00

1

 5

− 147,881

129

296,016

297,146

296,736

0.68

0.90

0.00

2

 6

−147,078

155

294,467

295,824

295,332

0.00

0.89

0.00

4

Level 1–4

 2

−133,617

51

267,337

267,783

267,620

0.00

0.90

0.00

0

 3

−130,798

77

261,750

262,423

262,178

0.00

0.89

0.00

1

 4

− 129,473

103

259,153

260,053

259,726

0.00

0.88

0.00

2

 5

−129,004

129

258,267

259,394

258,984

0.68

0.85

0.00

3

 6

− 128,604

155

257,519

258,874

258,381

0.10

0.86

0.00

5

Level 5

 2

−13,929

51

27,960

28,208

28,046

0.00

0.83

0.000

0

 3

−13,748

77

27,651

28,024

27,780

0.0004

0.85

0.000

0

 4

−13,579

103

27,365

27,864

27,537

0.001

0.87

0.000

0

 5

−13,456

129

27,170

27,795

27,386

0.11

0.83

0.000

0

 6

−13,389

155

27,088

27,839

27,347

0.72

0.78

0.000

1

  1. For binary responses, SCL-90-R (cutoff = 2); PHQ9 (cutoff = 2)
  2. LMRT Lo-Mendell-Rubin test
  3. BLRT Bootstrap Likelihood Ratio Test
  4. Numbers in bold are indicative when selecting the best model