The Least Square Dummy Variable (LSDV) method can be used to estimate parameters in the panel data regression model incomplete one-way fixed effect. To produce the best model with GDP data of GRASB. Variables that do not occur heteroscedasticity and models that meet the smallest sum square of error is the variable Mining and Processing Industry, this variable affects the per capita income. The Feasible Generalized Least Square (FGLS) method can be used to estimate the regression parameters for incomplete panel data for a one-way random effect. In this model produce the best model with non-oil and gas GRDP data. The variables that fulfill it are the processing Industry, service, and agriculture of Forestry and Fishery.  Therefore looking at the above model can be concluded non-oil and Gas GRDP has three factors that affect per capita income in Bangka Belitung. This should be a reference of local governments to further improve the quality or production in agriculture and services because this potential is more promising for the future. Software used to analyze data in this paper is with R.