In this research, heterogeneity of variance components in cattle populations at climatical regions in the first three lactations was studied. The data set included the following: 161328 records of first, 123369 records of second and 81013 records of third lactations, which was collected by Animal Breeding Center of Iran from 1983 to 2004. Records of three lactations were divided in the base of Domarten method. Bartlett test for heterogeneity of variance components was significant among all subgroups. In order to decrease the heterogeneity of variance components, we used several data transformation methods including Logarithmic, Square root and Arc sin transformations. Logarithmic transformation decreased the heterogeneity of variance components in the three lactations and other methods had not effect for removing the heterogeneity in any group. Genetic parameters and heritability were estimated for three lactations by MATVEC program, using animal model. Results showed that the heritability estimates of milk yield were decreased from the first lactation to the third; also the heritability estimates of transformed data were slightly higher than the original data. Comparison between estimated parameters in single trait and two traits analysis, before and after data transformation, showed that there were not significant differences between derived results.