On the efficiency of estimators in truncated height samples

We test the efficiency of estimators proposed for truncated height samples with a new data set of over 23,000 height observations covering nearly all conscripts in Drenthe, a province of the Netherlands, over the period 1826-1860. We find that the `best' estimator, truncated ML, in its unrestricted form overestimates the mean and underestimates the variance. If the variance is set to the population variance, the mean is underestimated. We question the normality assumption that is typically made in this literature. Our `population' is skewed, which might explain the poor performance of the estimators