Exploring the suitability of non-conventional method of imputation for compilation of CPI Sinha Dilip Kumar1,*, Saxena Mohita2 1Indian Statistical Service Officer, posted as a Deputy Director General in the Ministry of Statistics and Programme Implementation. He has working experience of twenty-two years in official statistics, including nine years in the field of price statistics 2Indian Statistical Service Officer, posted as an Assistant Director in the Ministry of Health and Family Welfare *Corresponding author email id: dilipkumar_sinha@yahoo.co.in
Online published on 22 March, 2024. Abstract We come across with various kind of data sets every day. These are found, inevitably, containing outliers and missing observations due to different types of non-sampling errors. The missing data introduce bias because of many reasons, including data from reduced sample compromising on representativeness. Therefore, it becomes essential to impute values for the noise and misreported data points. There are many well-known methods for imputation to deal with the aforementioned problems. We continuously strive for finding a better method of imputation which results in maximum reduction of bias in the estimates or indicators compiled. It is, therefore, the need of the hour to go beyond conventional method of imputation. In this paper, a non-conventional method has been explored in the field of price statistics for compilation of Consumer Price Index (CPI). Time Product Dummy (TPD) method has been found to be one of the best possible ways of imputing the missing observations. This method uses maximum information contained in the reported data and is based on spatial as well as temporal factors. TPD method can also be used for index compilation since this prepares index simultaneously while imputing the missing prices. It has been found that index from Jevon's method and that from the TPD method are exactly the same, once the tableau has been made complete after imputation using the later method. Top Keywords Compilation, Exploring, Imputation, Non-conventional, Suitability. Top |