Classical (parametric) regression techniques are based on the assumption that the independent variable is correlated linearly with the dependent variables and the pattern of this relation is known. When such assumption cannot be verified, parameter estimations fail to be reliable. In cases where the way of correlation is not known or it does not comply with the known parametric mathematical patterns, nonparametric regression techniques are to be applied. One shortcoming concerning this procedure emerges particularly in the interpretation process due to problems brought about by multidimensional aspect of the existence of more than one independent variable. Whenever confronted with a case that includes more than one independent variable, some of the independent variables correlate linearly with the dependent variable; at other times some of the independent variables might correlate nonlinearly. In order to establish a modeling for such relations, semiparametric regression models, comprising the aggregate of parametric and nonparametric regression function, are utilized. İn this study semiparametric regression definitions, estimation (backfitting algorithm), confidence bands, calculating standard errors and hypothesis testing are explained.