|If no information at all was available about the shape of a regression, then the averages of the observed values response variable Y for each value of the explaining variable X would have to suffice as estimates of the conditional expectations. On the other hand, if the regression-function was perfectly known except for some parameters, then the problem would be reduced to theat of estimating the parameter vector. In this project an intermediate situation with limited information about the shape is treated. The restrictions of the shape are qualitative and are combinations of properties such as: monotone, unimodal and convex.
Papers from the Statistical Research Unit, University of Gothenburg
Frisén, M. (2011) Detection of turning points in business cycles. International encyclopedia of statistical sciences, Ed. Lovric, M. Springer.
Frisén, M., Andersson, E. and Pettersson, K. (2010) Semiparametric estimation of outbreak regression. Statistics 44, 107-117.
On Unimodal Regression in the Exponential Family. Communications in Statistics ¿ Theory and Methods, 38, 1526-1538.