Presentation
I. Overview
FOCUSHow ML methods
work
Two alternative ML
approaches
Guidelines for choice
of ML approach
Overview of
inferencesII. Background About
Maximum Likelihood
Procedure
Maximum likelihood (ML)
estimation
Least squares (LS) estimation: used
in classical linear regression
ML¼LS when normality is
assumed
ML estimation:
Computer programs available
General applicability
Used for nonlinear models, e.g.,
the logistic model
This presentation gives an overview of maxi-
mum likelihood (ML) methods as used in logis-
tic regression analysis. We focus on how ML
methods work, we distinguish between two
alternative ML approaches, and we give guide-
lines regarding which approach to choose. We
also give a brief overview on making statistical
inferences using ML techniques.Maximum likelihood (ML)estimation is one
of several alternative approaches that statisti-
cians have developed for estimating the para-
meters in a mathematical model. Another
well-known and popular approach is least
squares (LS) estimation which is described
in most introductory statistics courses as a
method for estimating the parameters in a
classical straight line or multiple linear regres-
sion model. ML estimation and least squares
estimation are different approaches that hap-
pen to give the same results for classical linear
regression analyses when the dependent vari-
able is assumed to be normally distributed.For many years, ML estimation was not widely
used because no computer software programs
were available to carry out the complex calcu-
lations required. However, ML programs have
been widely available in recent years. More-
over, when compared with least squares, the
ML method can be applied in the estimation
of complex nonlinear as well as linear models.
In particular, because the logistic model is a
nonlinear model, ML estimation is the preferred
estimation method for logistic regression.106 4. Maximum Likelihood Techniques: An Overview