I comment here on a recent paper in this journal, on the fitting of truncated normal distributions by the EM algorithm. I show that the fitting of such distributions by direct numerical maximization of likelihood (rather than EM) is straightforward, contrary to an assertion made by the authors of that paper.
TianGLJuDYuenKCet al.New expectation–maximization-type algorithms via stochastic representation for the analysis of truncated normal data with applications in biomedicine. Stat Meth Med Res. Epub ahead of print 13 December 2016. DOI: 10.1177/0962280216681598.
2.
CohenACJr. Estimating the mean and variance of normal populations from singly truncated and doubly truncated samples. Ann Math Stat1950; 21: 557–569.
3.
CohenACJr. Simplified estimators for the normal distribution when samples are singly censored or truncated. Technometrics1959; 1: 217–237.
4.
R Core Team. R: A Language and Environment for Statistical Computing. Austria: R Foundation for Statistical Computing, Vienna, www.R-project.org/ (2014, accessed 13 May 2017).
5.
MacDonaldIL. Numerical maximisation of likelihood: a neglected alternative to EM?Int Stat Rev2014; 82: 296–308.
6.
SpäthH. Mathematical algorithms for linear regression, San Diego, CA: Academic Press, 1992.
7.
KleinbaumDGKupperLLNizamAet al.Applied regression analysis and other multivariable methods, 5th ed. Boston, MA: Cengage Learning, 2014.
