We introduce a method that integrates causal inference with multi-level modeling (MCBNS) to develop a prediction model grounded in stable causal relationships and refine it at a fine-grained scale.
bnlearn_pop_R20123c2020_same2020.R, bnlearn_pop_R20123c2010_same2020.R, bnlearn_pop_R20123c2000_same2020.R: Causal model prediction
bnlearn_lme5_2020_shen.R, bnlearn_lme5_2010_shen.R, bnlearn_lme5_2000_shen.R: Multiscale correction
city_pop_Y6_Lat_geo9.txt: 2020 County-Level Population Data and Influencing Factors, twon_pop2020_all_class7.txt: 2020 Population Data for Townships and Influencing Factors, China1000Asia_Latp_bf4_geo10_1.txt: 2020 Population Projections by Kilometer Grid, pop_predict_bnlearn_Lat55f_4.txt: Linking 2020 population projections for the kilometer grid with township data;
city_pop_Y6_Lat_geo9_2010.txt: 2010 County-Level Population Data and Influencing Factors, twon_pop2010_all_class7.txt: 2010 Population Data for Townships and Influencing Factors, China1000Asia_Latp_bf4_geo9_2010.txt: 2010 Population Projections by Kilometer Grid, pop_predict_bnlearn_Lat55f_2010_2.txt: Linking 2010 population projections for the kilometer grid with township data;
city_pop_Y6_Lat_geo9_2000.txt: 2000 County-Level Population Data and Influencing Factors, twon_pop2000_all_class7.txt: 2000 Population Data for Townships and Influencing Factors, China1000Asia_Latp_bf4_geo9_2000.txt: 2000 Population Projections by Kilometer Grid, pop_predict_bnlearn_Lat55f_2000_2.txt: Linking 2000 population projections for the kilometer grid with township data.
Because some of the training data files are too large, the larger files have been uploaded to Figshare. The complete dataset is available on Figshare (https://doi.org/10.6084/m9.figshare.31850800).