This data set has been derived from the Quarterly Interview Survey of the Consumer Expenditure Survey (CES) undertaken by the U.S. Department of Labor, Bureau of Labor Statistics and is available at https://www.bls.gov/cex/ where also more details about this survey can be found. The original data set comprises 869 households in 34 variables of which one is unique ID, five characterize the size of the household, further 6 variables contain other characteristics of the household like age, education ethnicity, etc. and 22 variables represent the household expenditures. We will consider a reduced set of only 8 expendature variables. This reduced data set was analyzed by Hubert at al. (2009)in the context of PCA and the first step of the analysis showed that all variables are highly skewed. They applied the robust PCA method of Serneels and Verdonck based on the EM algorithm, since some of the data are incomplete.
data(ces)A data frame with 869 observations on the following 8 variables:
EXPTotal household expenditure
FDHOFood and nonalcoholic beverages consumed at home
FDAWFood and nonalcoholic beverages consumed away from home
SHELHousing expenditure
TELETelephone services
CLOTClothing
HEALHealth care
ENTEntertainment
Hubert, M, Rousseeuw, P.J. and Verdonck, T., (2009). Robust PCA for skewed data and its outlier map, Computational Statistics & Data Analysis, 53, 6, pp. 2264-2274
data(ces)
summary(ces)
plot(ces)
Run the code above in your browser using DataCamp Workspace