Assessing Robustness of Regression Results in Sequence Analysis
Author Information
Author(s): Leonard Roth, Matthias Studer, Emilie Zuercher, Isabelle Peytremann-Bridevaux
Primary Institution: University of Lausanne
Hypothesis
How does sampling uncertainty affect the robustness of regression results in sequence analysis?
Conclusion
The proposed method reveals that the original effect of regular lipid testing on healthcare utilization patterns was overestimated due to inadequate handling of sampling error.
Supporting Evidence
- The bootstrap procedure produced an average marginal effect matrix for the association between regular lipid testing and cluster membership.
- High cluster-wise stability was observed with Jaccard coefficients above 0.9.
- The pooled average marginal effect indicated a reduced association when accounting for sampling uncertainty.
Takeaway
This study shows that when researchers analyze patterns in healthcare use, they need to be careful about how they handle data uncertainty, or they might think something is true when it isn't.
Methodology
The study used a bootstrap procedure to assess the robustness of regression results from sequence analysis by creating multiple samples and evaluating the stability of clustering and regression outcomes.
Potential Biases
Self-reported data may introduce recall bias, and the sample may not fully represent the population due to attrition.
Limitations
The study's findings may be influenced by a narrow study period, reliance on self-reported data, and potential attrition bias.
Participant Demographics
The study included 348 non-institutionalized adult patients with diabetes, primarily from the canton of Vaud, Switzerland.
Statistical Information
P-Value
0.271
Confidence Interval
[−0.016, 0.482]
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website