Table 5 The MN and MN-MS generally predict out-of-sample mosquito data better than other competing regression models.

From: Ordinal regression models for zero-inflated and/or over-dispersed count data

Species

Predictive performance

MN model

MN-MS model

Poisson

NB

ZINB

ZIP

Poisson

NB

ZINB

ZIP

MN

A. darlingi

0.86

0.79

0.79

0.64

0.86

0.79

0.79

0.64

0.36

A. nuneztovari

0.86

0.86

0.93

1.00

0.93

0.79

0.93

1.00

0.57

A. triannulatus

0.79

0.71

0.79

0.64

0.79

0.71

0.79

0.64

0.29

A. benarrochi

0.79

0.79

0.86

0.71

0.79

0.86

0.93

0.71

0.79

A. oswaldoi

0.79

0.86

0.71

0.71

0.79

0.79

0.79

0.79

0.64

A. rangeli

0.79

0.93

0.93

0.79

0.71

0.93

0.93

0.79

0.79

  1. Numbers indicate the proportion of cross-validation folds (based on 14 folds) in which the MN and MN-MS models had lower MSE scores when compared to each alternative model and for each mosquito species. “ZI” stands for zero-inflation. The last column on the right shows the proportion of cross-validation folds in which the MN-MS model had lower MSE score relative to the MN model.
Back to article page