Bibliographic citations
Collazos, O., (2021). Imputación de datos faltantes en los ingresos por hogar en la Enaho utilizando el método del K-vecino más cercano [Universidad Nacional Agraria La Molina]. https://hdl.handle.net/20.500.12996/4851
Collazos, O., Imputación de datos faltantes en los ingresos por hogar en la Enaho utilizando el método del K-vecino más cercano []. PE: Universidad Nacional Agraria La Molina; 2021. https://hdl.handle.net/20.500.12996/4851
@misc{renati/249123,
title = "Imputación de datos faltantes en los ingresos por hogar en la Enaho utilizando el método del K-vecino más cercano",
author = "Collazos Tuesta, Oscar Ronald",
publisher = "Universidad Nacional Agraria La Molina",
year = "2021"
}
The National Household Survey (ENAHO) is the instrument used by the National Institute of Statistics and Informatics (INEI) to collect national data on household economic, educational and health conditions, etc. and that allow generating indicators that measure the status and evolution of poverty, well-being and living conditions of Peruvian households, as well as to carry out diagnoses and measure the scope of social programs (food and non-food) in the improvement of the living conditions of the Peruvian population. However, a problem that ENAHO must face is the total or partial non-response in the sampling units (non-response in units) or in a specific question (non-response per item); especially to the questions referring to the income of the households. For the treatment of missing data, a variety of methods have been proposed , ranging from the simplest, which consists of elimination of observations that have some missing data in one of the variables, to most consistent methods based on an imputation process with the missing data from the complete data. The objective of this research is to present and apply the imputation methods of the mean and median, the Hot-Deck method and the nearest k neighbor to estimate the missing data of the Income per household in the ENAHO 2017 quarter 3. The results indicate that missing income data has a MCAR mechanism. The estimate of the 95% confidence interval for the mean of the imputed income, had amplitudes by the method of the mean 131.41 (the smallest) while for the nearest k neighbor it was 139.4. To estimate the standard deviation of income, it was the lowest for the mean 92.97 and k nearest neighbor 100.99. The results of the comparison of the imputation methods, were using the complete data to generate a random sample of artificial missing data, and then the Mean Square Error (ECM) and correlations with the observed and imputed data for each method were found. The closest neighbor k method had the lowest ECM values of 1412.6 and 444.4 for the mean and median; while the other methods their values were by the average 1504.5; by the median 1619.9 and by the Hot-Deck 1963.7. The correlation coefficients resulted in very similar values, for k nearest neighbor 0.968 with the mean and 0.964 with the median.
This item is licensed under a Creative Commons License