National surveys are often representative at the national level but not at a regional level which makes estimation of sub-national/regional public opinion challenging. Multilevel regression with post-stratification is the current gold standard for this type of problem and autoMrP implements this approach in R. In addition, autoMrP applies machine learning techniques to further improve the prediction performance of this approach. For more information please refer to the final pre-print version of Broniecki, Wüest, Leemann (2020) "Improving Multilevel Regression with Post-Stratification Through Machine Learning (autoMrP)" forthcoming in the Journal of Politics (https://lucasleemann.files.wordpress.com/2020/07/automrp-r2pa.pdf).
Philipp Broniecki has written a tutorial for the package.