Cumulative logit modelling for ordinal response variables: applications to biomedical research

Area of Science:

• Biostatistics
• Epidemiology
• Medical Research

Background:

• Ordinal response data analysis frequently employs inappropriate statistical methods.
• Common practices like using mean scores or Pearson chi-square tests are fallacious as they ignore data's inherent ranking and equidistant properties.
• Existing non-parametric methods may consider ordering but often fail to adjust for confounding or assess effect modification.

Purpose of the Study:

• To highlight the inadequacy of current statistical methods for ordinal data analysis.
• To introduce and advocate for the cumulative logit model as a suitable alternative.
• To demonstrate the practical application of the cumulative logit model in biological and medical research.

Main Methods:

• The study critiques common statistical methods (mean scores, Pearson chi-square) for ordinal data.
• It introduces the cumulative logit model, a multivariate statistical technique.
• The model's ability to handle ranked data, adjust for confounding, and assess effect modification is emphasized, supported by research examples.

Main Results:

• The cumulative logit model effectively analyzes ordinal response data by respecting inherent order.
• This multivariate approach allows for statistical adjustment of confounding variables.
• It also facilitates the assessment of effect modification, even with modest sample sizes.

Conclusions:

• The cumulative logit model is a powerful and appropriate tool for analyzing ordinal response data in etiologic inference.
• It overcomes limitations of traditional methods by incorporating data ranking and enabling complexAdjustments.
• The model supports robust analysis essential for biological and medical research, with SAS programs available for implementation.