A Power Transformation Method for Incomplete Two Way Contingency Tables with Nonignorable Nonrespose
Seongyong Kim, Boseung Choi, Yousung Park
Economics and Statistics Institute, Korea University Sejong Campus, Yeonki-Gun, Chungnam, Republic of Korea; Department of Computer Science and Statistics, Daegu University, Gyeongsan, Gyeongbuk, Republic of Korea; Department of Statistics, Korea University, Seoul, Republic of Korea

For the analysis of the incomplete contingency tables with nonignorable nonresponse, the maximum likelihood estimates (ML) often give rise to estimates on the boundary solution of the parameter space and to imperfect fits for saturated log linear models. We present the explicit condition for the boundary solution arising from ML estimation for contingency tables with nonignorable nonresponse. To escape those conditions, we propose to use a power transformation as a link function instead of the usual logit link function, and show a power link function performed better than a logit link function in both ML estimation and Bayesian estimation under several scenarios of simulated data sets. We also demonstrate that a saturated nonignorable nonresponse model imperfectly fits the observation because the correlations of estimates are close to 1.

References:

Baker, S. G., and Laird, N. M. (1988). Regression analysis for categorical vaiables with outcome subject to nonignorable nonresponse. Journal of the American Statistical Association, 83, 62-69.

Baker, S. G., Rosenberger, W. F. and Dersimonian, R. (1992). Closed-form estimates for missing counts in two-way contingency tables. Statistics in Medicine, 11, 643-657.

Choi, B. S., Choi, J. W. and Park, Y. (2009). Bayesian methods for an incomplete two-way contingency table with application to the Ohio (Buckeye State) Polls. Survey Methodology, 35, 37-51.

Park, T., and Brown, M. B. (1994). Models for categorical data with nonignorable nonresponse. Journal of the American Statistical Association, 89, 44-52.

Park, T. (1998). An approach to categorical data with nonignorable nonresponse. Biometrics, 54, 1579-1690.

Park, Y., and Choi, B. S. (2010). Bayesian analysis for incomplete multi-way contingency tables with nonignorable nonresponse. Journal of Applied Statistics, 37, 1439-1453.

Keywords: Boundary solutions problem; Generalized Linear Model; ML estimation; Bayesian estimation

Biography: Seongyong Kim, a Ph. D in Statistics since 2011, now works for Economics and Statistics Institute, Korea University Sejong Campus as a Research Professor.