Poisson Regression
The high-fertility tail was much heavier than Poisson expectations. The observed child-count distribution had too many zero-birth women and too few women with one through four births relative to a matched Poisson curve. The child-count dis…
1 sources - 5 claims
The high-fertility tail was much heavier than Poisson expectations. The observed child-count distribution had too many zero-birth women and too few women with one through four births relative to a matched Poisson curve. The child-count distribution in the peerage data is not Poisson. The paper says the Poisson regression method is valid as a trend indicator only if child counts follow a Poisson distribution. Removing nine women with the most children caused Poisson regression to stop producing a negative coefficient.