Seeing the forest despite the trees. The benefit of exploratory data analysis to program evaluation research

Area of Science:

• Rehabilitation Medicine
• Health Services Research
• Biostatistics

Background:

• Program evaluation often relies on conventional statistical methods.
• The validity of these methods can be compromised by non-normal data distributions.
• Exploratory Data Analysis (EDA) offers alternative approaches for data interpretation.

Purpose of the Study:

• To demonstrate the benefits of applying EDA techniques to program evaluation.
• To evaluate an intensive rehabilitation program for rheumatoid arthritis patients.
• To compare the effectiveness of intensive rehabilitation versus customary care.

Main Methods:

• Comparison of perceived health status between intensive rehabilitation and customary care groups over 18 months.
• Conventional analysis using analysis of variance (ANOVA).
• Exploratory data analysis using graphic displays (medians and boxplots).

Main Results:

• Conventional analysis indicated overall patient improvement with no significant difference between groups.
• EDA revealed non-normal outcome variable distribution, questioning conventional analysis validity.
• EDA showed the intensive rehabilitation group initially lagged but showed significant improvement at 18 months.

Conclusions:

• EDA provides critical insights beyond conventional statistical methods in program evaluation.
• The delayed improvement in the intensive rehabilitation group suggests potential selection bias or increased patient awareness.
• EDA enhances the understanding of complex data patterns in health services research.