|
Number of Classes
|
Number of parameters
|
Log Likelihood
|
Bootstrapped p-value of L2a
|
BIC (LL)b
|
Classification Errorc
|
Entropy R-squaredd
|
|---|
|
1
|
32
|
-5,565
|
0.084
|
11,337
|
---
|
---
|
|
2
|
42
|
-5,445
|
0.078
|
11,162
|
0.099
|
0.635
|
|
3
|
52
|
-5,394
|
0.204
|
11,125
|
0.156
|
0.647
|
|
4
|
62
|
-5,381
|
0.138
|
11,163
|
0.190
|
0.629
|
|
5
|
72
|
-5,367
|
0.128
|
11,201
|
0.220
|
0.624
|
-
aThe bootstrapped p-value of the likelihood-ratio statistic indicates if the model-based estimated frequencies are of sufficient agreement to observed frequencies, i.e. the extent to which the model fits the data, with p < 0.05 indicating a poor fit. Bootstrapping is used as the chi-squared distribution may not provide a good approximation to the distribution of the statistic, and hence provides more valid estimates. Bolded values indicate estimate of best fit.
-
bBIC adjusts the LL value for the number of parameters in the model, thus accounting for model parsimony, with a lower value indicating a more preferable model.
-
cClassification error indicates the proportion of cases that are estimated to be misclassified when cases are classified to the class for which they have the highest posterior probability, with values closer to 0 desirable.
-
dEntropy R-squared value indicates how well class membership can be predicted based on the indicator variables , with values closer to 1 desirable.
