CM6.{3,5-6} | CM6.{3,5-6} | Statistical Analysis and Software Use — Summary & Reflection

KEY TAKEAWAYS

This module has equipped you with the inferential tools that transform descriptive data into evidence-based decisions in community medicine:

1. Inference and error (CM6.3): Every statistical conclusion operates under uncertainty. The null hypothesis framework, Type I error (α = 0.05 false positive), and Type II error (β; power = 1 − β) define the decision framework. p < 0.05 is a convention, not an absolute truth criterion.

1. Normal distribution and confidence intervals (CM6.3): The 68-95-99.7 rule, z-scores, and the Central Limit Theorem justify parametric methods for large samples. The 95% CI = mean ± 1.96 × SE; a CI that excludes the null value indicates statistical significance.

1. Test selection decision framework (CM6.3): The sequential decision — data type → normality → groups → pairing — determines the correct test: z-test (large n, known σ); t-tests (three forms: one-sample, independent, paired); ANOVA (≥3 groups); chi-square/Fisher's (categorical); Mann-Whitney/Kruskal-Wallis (non-parametric).

1. Interpretation (CM6.3, CM6.6): p-value = probability of data under H₀ (not probability H₀ is true). Statistical significance ≠ clinical significance. Report exact p-values, effect sizes, and CIs. Beware multiple comparisons and expected-frequency requirements for chi-square.

1. Software (CM6.5): Epi Info (field epidemiology, free), OpenEpi (browser-based, sample size and chi-square), SPSS (research and thesis), Excel (basic calculations). Know the function of each for NMC examinations and community posting practice.

1. Applied practice (CM6.6): The four-step workflow — descriptive summary, graphical display, appropriate significance test with verification of conditions, careful contextualised interpretation — applies to any community medicine dataset.

REFLECT

Return to the hook: the health officer evaluating the iron-fortified rice programme, where haemoglobin was 10.9 g/dL in programme villages vs 10.3 g/dL in control villages. Using what you now know: What additional information would you need before selecting a statistical test? (Think: sample size, whether the same villages were measured before and after vs two different groups, whether haemoglobin is normally distributed.) If the officer reports p = 0.03 and you are advising the State government on a ₹45-crore scale-up decision, what other statistical and programmatic information would you want before recommending scale-up? Is a 0.6 g/dL difference clinically meaningful for the population this programme targets? Write your reasoning in 150 words, then compare with a classmate.