Closed solivella closed 2 years ago
For surname only, could you comment on why we think it's reasonable to redistribute using a fixed race marginal? Does this not defeat the intuition that race will be distributed based on the most likely race by surname when surname.only = TRUE
?
For surname only, could you comment on why we think it's reasonable to redistribute using a fixed race marginal? Does this not defeat the intuition that race will be distributed based on the most likely race by surname when
surname.only = TRUE
?
surname.only = TRUE
computes Pr(race | surname)
, but our current name dictionaries return Pr(surname | race)
. To "invert" the probability, we need to multiply Pr(surname | race)
by Pr(race)
(and then normalize), as per Bayes' rule.
Resolve issue reported in #68.