Demonstration of an Exploratory Method for Categorical Data Imputing Inventories Zero or Non-zero Values

Anri Mutoh (
Rissho University, Tokyo, Japan
Ichiro Murata (
National Statistics Center, Tokyo, Japan


In Unincorporated Enterprise Survey of Japan, the problem of poor accuracy in the imputation of missing values for Inventories was caused by the inclusion of many zero values. To address this issue, we used a strategy that involves identifying categorical variables at first that could potentially indicate whether Inventories are zero or non-zero. This was done through summary statistics S1 which was obtained through Exploratory Data Analysis. S1 is simply a summary statistic and is not designed to accurately predict the values of Inventories. Therefore, we conducted an experiment using real data to test whether S1 could guarantee the same accuracy as other methods that pursue higher accuracy in variable selection. The results showed that S1 was as accurate as these other methods. Additionally, the variables selected by the other methods were difficult to interpret, but the variables chosen by S1 were easily understandable based on practical experience.
Keywords: Unincorporated Enterprise Survey, categorical data analysis, inventories, Exploratory Data Analysis
JEL Classification: C14

[Full Text]

Romanian Statistical Review 1/2023