An important and very general problem!
One possibility is to use ideas developed to deal with missing data in statistics.
That is to impute missing values using some systematic approach, and to use multiple imputations to summarize the extra uncertainty that brings to the analysis.
One possibility is to use ideas developed to deal with missing data in statistics.
That is to impute missing values using some systematic approach, and to use multiple imputations to summarize the extra uncertainty that brings to the analysis.
Comments
About 20-years ago I wrote a couple of articles attempting to translate those ideas to historical cases:
https://direct.mit.edu/jinh/article-abstract/36/2/209/48104/Propensity-Score-Methods-and-the-Lenin-School
I still think that is the right basic approach. However, to be honest the implementation isn't that great, and I'd do thing rather differently if I were doing something similar now.