Skip to main content

Table 3 Determination of the optimal number of segments

From: Consumer demand heterogeneity and valuation of value-added pulse products: a case of precooked beans in Uganda

No. of segments

P

ρ 2

LL

AIC

BIC

p-value**

1

6

0.17

3210.53

6433.1

3235.35

<.001

2

21

0.32

2907.61

5857.2

2994.45

<.001

3

36

0.35

2738.23

5548.5

2887.09

<.001

4

51

0.30

2963.71

6029.4

3174.59

<.001

  1. Sample size 3906 choices from 558 consumers, number of observations = 11,718. Equations: ρ2 = 1 − (LL)/LL (o); AIC = − 2LL + 3P and BIC = − LL + (P/2) × ln (N). The LL is a negative number. ** Log-likelihood ratio test (Prob. chi squared value)
  2. In terms of the Log-likelihood, all models tested (a model with 1, 2, 3 and 4 segment(s)) had the same p-value of 0.000. The values of the LL, AIC, BIC, and ρ2 were then used in model selection from which the three-segment met the most required criteria
  3. Italic values indicate a model with the best LL and combination of information criteria