Logistic regression is a statistical means of creating a prediction function based on a sample. This form of regression is most often used by researchers who suspect that the outcome of a situation is not linearly related to the independent variables of the study. Logistic regression is thus an alternative to linear regression, based on the "logit" function, which is a ratio of the odds of success to the odds of failure. However, despite its lack of need for reliance on assumptions of linearity, logistic regression has its own assumptions and traits that make it disadvantageous in certain situations.

1

Restrictions on the Dependent Variable

Unlike linear regression, logistic regression can only be used to predict discrete functions. Therefore, the dependent variable of logistic regression is restricted to the discrete number set. This restriction itself is problematic, as it is prohibitive to the prediction of continuous data. An addition problem with this trait of logistic regression is that because the logit function itself is continuous, some users of logistic regression may misunderstand, believing that logistic regression can be applied to continuous variables.

  • Unlike linear regression, logistic regression can only be used to predict discrete functions.
  • An addition problem with this trait of logistic regression is that because the logit function itself is continuous, some users of logistic regression may misunderstand, believing that logistic regression can be applied to continuous variables.
2

Large Sample Size

Logistic regression can accept a large number of independent variables. While this may seem like an advantage, there are many situations when it is not. Because the parameter estimation procedure of logistic regression relies heavily on having an adequate number of samples for each combination of independent variables, small sample sizes can lead to widely inaccurate estimates of parameters. Thus, users of logistic regression should first make sure they can obtain a sample of large size before deciding on logistic regression as the analysis method.

  • Logistic regression can accept a large number of independent variables.
  • Because the parameter estimation procedure of logistic regression relies heavily on having an adequate number of samples for each combination of independent variables, small sample sizes can lead to widely inaccurate estimates of parameters.
3

Assumption of Linearity

A researcher discarding linear regression models in favour of logistic regression models is likely doing so because the assumption of linearity between the dependent variable and the independent variables is unreasonable. However, what many researchers do not realise is that logistic regression also has an implicit assumption of linearity in terms of the logit function versus the independent variables. This assumption is fairly unreasonable as well.

4

Only for Between-Subject Designs

Logistic regression can only apply to studies using between-subject designs. This means within-subject designs preclude logistic regression methods. In many forms of research, especially those using human subjects, within-subject designs are preferred, as they can conserve resources. Thus, while in the fields of medicine and psychology logistic regression may seem suitable, in fact it cannot always be a choice.

  • Logistic regression can only apply to studies using between-subject designs.
  • In many forms of research, especially those using human subjects, within-subject designs are preferred, as they can conserve resources.