11.8. Summary

Statistical analysis in psychological research bridges sample data and broader population inferences, utilising descriptive measures like means or correlation coefficients to estimate parameters. Sampling variability introduces potential errors, which null hypothesis significance testing (NHST) addresses by determining whether observed patterns reflect true population relationships or random chance. NHST evaluates the null hypothesis (no relationship) against the alternative hypothesis (a genuine relationship), using the p-value to guide conclusions. A p-value below 0.05 typically indicates statistical significance, but researchers must distinguish this from practical or clinical significance, which gauges real-world relevance. Tools like t-tests and analysis of variance (ANOVA) provide foundational methods for comparing means and exploring group differences, while Pearson’s r quantifies relationships between variables, requiring significance testing to validate findings.

Errors like Type I (false positives) and Type II (false negatives) can undermine conclusions, emphasising the need for robust designs and appropriate statistical power, typically set at 0.80 to balance error minimisation and resource efficiency. Low power, often due to small sample sizes or weak effects, risks unreliable results but can be mitigated through larger samples, stronger manipulations, or within-subjects designs. Additionally, challenges like publication bias and p-hacking distort research integrity, necessitating solutions such as replication, registered reports, and transparent data sharing to ensure unbiased findings.

Despite its centrality, NHST faces criticisms for misinterpretations, particularly around p-values, and for its reliance on rigid significance thresholds like p < 0.05, which often dismiss nearly significant results and underreport non-significant findings. Critics also argue that NHST provides limited insights into effect strength or practical implications, suggesting alternatives like Bayesian statistics or complementary practices such as reporting effect sizes and confidence intervals. These enhancements offer richer insights and address the limitations of NHST, which remains widely used but is increasingly supplemented by innovative approaches to ensure rigorous and meaningful research outcome