Abstract:
This study attempts to apply ordinal logistic regression to identify the determinants of women
nutritional status using the data of data of Ethiopian Demographic and Health survey 2011.
Based on body mass index nutritional status is categorized in to four groups--severely under
nutrition, moderately under nutrition, normal and over\excess nutrition. Since nutritional status is
ordinal, an ordinal logistic regression model can be developed instead of three separate binary
logistic regression model to find predictors of nutritional status if the proportional odd
assumption satisfies. The assumption is violated with two predictors so non-proportional odd
model, Fienberg‘s continuation model without proportional odds, adjacent categories, and
separate binary logistic regression have been developed.
The model determines household wealth status and marital status were the significant predictors
of women nutritional status
The findings clearly justify that ordinal logistic regression model, cumulative logistic model with
non-proportional odds are more appropriate model for fitting nutritional status of women than
Fienberg‘s continuation ratio and adjacent category models and marital status and wealth status
are significant variables.