Skripsi
PENERAPAN METODE CHI-SQUARE AUTOMATIC INTERACTION DETECTION (CHAID) DAN MODEL REGRESI LOGISTIK ORDINAL PADA TINGKAT PREVALENSI KETIDAKCUKUPAN KONSUMSI PANGAN DI INDONESIA
This study aims to explore factors closely associated with the Prevalence of Undernourishment (PoU) in Indonesia and to estimate the probability of PoU levels based on influential variables. The methods used are CHAID and ordinal logistic regression. The independent variables analyzed include rice production, population density, percentage of poor population, percentage of per capita food expenditure, Desirable Dietary Pattern (DDP), Human Development Index (HDI), and Food Security Index (FSI). The data were obtained from the official websites of Statistics Indonesia (BPS) and the National Food Agency (NFA). The CHAID exploration results show that the six variables most strongly associated with PoU in order are FSI, percentage of poor population, population density, HDI, DDP, and percentage of per capita food expenditure. Three variables, namely percentage of poor population, DDP, and percentage of per capita food expenditure, experienced category merging. The recategorized variables based on the CHAID results were subsequently analyzed using an ordinal logistic regression model. The results of the best ordinal logistic regression model estimation show that four variables have a significant effect on PoU levels, namely population density, percentage of poor population, DDP, and FSI, with an accuracy rate of 82.3 percent. The Odds Ratio indicates that regions with a vulnerable FSI have a 10.432 times greater probability of being at a higher PoU level compared to regions with a highly food secure FSI. In other words, regions with a vulnerable FSI tend to have worse PoU levels, therefore policies to strengthen food security should be focused on those regions.