The age, period and cohort (APC) model has received a considerable popularity for statistical analysis of data on consumer behaviors over the long term. In APC model, however, there exists an exact linear dependency among the three factors; hence even the first-order (linear) trend of each factor cannot be identified in the model. In this study, as an alternative one to APC model, a model which has age, period and environmental (APE) effects as its parameters is introduced for analyzing (age, period)-tabulated data on consumer survey for marketing research. It is shown that APE model is free from the non-identifiability problem from which APC model suffers though it is equivalent to APC model in terms of space spanned by column vectors in its design matrix. Further the results of fitting APE model to the data on consumer preference obtained from JNN Data Bank conducted by Japanese 28 TV stations including Tokyo Broadcasting System, Inc. (TBS) as their key station are shown. It is figured out, from those results, how Japanese food culture has been changed in the past several decades with respect to age, period and environmental (social) factors.
Hanayama, N. (2001). A simple two-stage model for cancer risk in the environment. Environmetrics 12, 757-773.
Hanayama, N. (2004). Age-environment model for breast cancer. Environmetrics 15, 219-232.
Hanayama, N. (2007). An extended age period cohort model for analysing (age, period)-tabulated data. Statistics in Medicine 26, 3459-3475.
Hanayama, N. (2008). Age Period Cumulative Environment Model for Analyzing (Age, Period)-Tabulated Data on Breast Cancer Deaths. Communications in Statistics – Theory and Methods 37, 1770–1782.
Keywords: Age, period and cohort analysis for marketing research; Consumer preference; Identifiability problem; (age, period)-tabulated data
Biography: Nobutane Hanayama is an associate professor in Department of Information Technology at Shobi University in Japan. Here he is teaching “data analysis” in his classes. So far he has his manuscripts regarding statistical models for the analysis of environmental risks for cancer published in several academic journals including “Environmetrics” and “Statistics in Medicine”. Recently he is undertaking a consumer survey data analysis for marketing research.