Related Experiment Videos
["P" as in Program Package. About torturing data and significant "fishing expeditions"]
1Institutionen för informationsvetenskap (statistik), Uppsala universitet. adam.taube@dis.uu.se
Lakartidningen
|October 5, 2002
Summary
Statistical analysis of prognostic factors in medical research requires careful validation. Findings from small patient groups with many variables need verification in independent studies to avoid spurious significance and ensure reliability.
Area of Science:
- Biostatistics
- Medical Research Methodology
- Prognostic Factor Analysis
Context:
- Medical researchers utilize advanced statistical techniques for prognostic factor studies.
- Traditional analysis involves small patient cohorts with numerous explanatory variables.
- The practice of 'statistical fishing expeditions' using p-values is common.
Purpose:
- To highlight the limitations of traditional prognostic factor analysis in small datasets.
- To address the risks of spurious significance and non-significant relevant factors.
- To emphasize the critical need for independent validation of prognostic findings.
Summary:
- Small sample sizes with many variables can lead to non-significant relevant factors and spurious significances.
- Over-reliance on p-values for variable selection increases the risk of false positives.
- Prognostic factor efficacy must be confirmed in independent patient cohorts.
Impact:
- Ensures more reliable and generalizable prognostic factor identification in medical research.
- Improves the robustness of statistical findings, reducing the impact of chance associations.
- Promotes rigorous scientific validation, leading to better clinical decision-making and patient outcomes.