Statistics is where most medical research projects get stuck. You've collected your data, but now what? Which test should you use? What does the p-value really mean?
This guide cuts through the jargon and gives you practical guidance on statistical analysis for your thesis or publication.
Understanding the Basics
Variables: The Building Blocks
- Categorical (Qualitative): Gender, blood group, disease stage
- Numerical (Quantitative): Age, blood pressure, hemoglobin
- Ordinal: Pain scale (mild/moderate/severe), staging
The Famous p-value
The p-value tells you the probability of getting your results if there's actually no real effect. By convention:
- p < 0.05: Statistically significant
- p < 0.01: Highly significant
- p ≥ 0.05: Not statistically significant
Important: Statistical significance ≠ Clinical significance. A p-value of 0.001 doesn't mean the finding is clinically important. Always consider effect size and clinical relevance.
Choosing the Right Test
This is where most people struggle. Here's a simplified decision guide:
| What you're comparing | Type of data | Test to use |
|---|---|---|
| Two groups | Numerical (normal) | Independent t-test |
| Two groups | Numerical (not normal) | Mann-Whitney U |
| Same group, before/after | Numerical (normal) | Paired t-test |
| Same group, before/after | Numerical (not normal) | Wilcoxon signed-rank |
| Three+ groups | Numerical (normal) | ANOVA |
| Three+ groups | Numerical (not normal) | Kruskal-Wallis |
| Two categorical variables | Categorical | Chi-square test |
| Small sample categorical | Categorical | Fisher's exact test |
| Relationship between variables | Numerical | Pearson/Spearman correlation |
Sample Size Calculation
One of the most common reasons for thesis rejection: inadequate sample size.
You need to determine sample size BEFORE starting data collection. Factors that affect it:
- Effect size: How big a difference you expect to find
- Alpha (α): Usually 0.05 (5% chance of false positive)
- Power: Usually 80% (80% chance of detecting true effect)
- Variance: How spread out your data is expected to be
Free calculators: G*Power (download) or online calculators like OpenEpi.
Normality Testing
Before choosing between parametric (t-test, ANOVA) and non-parametric tests, check if your data is normally distributed:
- Shapiro-Wilk test: Best for small samples (n < 50)
- Kolmogorov-Smirnov: For larger samples
- Visual: Histogram, Q-Q plot
Common Statistical Software
- SPSS: User-friendly, widely used in medical research
- R: Free, powerful, steep learning curve
- Stata: Excellent for epidemiology
- GraphPad Prism: Great for graphs, limited analysis
- Excel: Basic analysis only, not recommended for publications
Presenting Results
For Continuous Data
- Normal distribution: Mean ± SD
- Skewed distribution: Median (IQR)
For Categorical Data
- Number and percentage: n (%)
Tables and Figures
- Tables should be self-explanatory with clear headers
- Include p-values and confidence intervals where appropriate
- Graphs should be clean, properly labeled
Need Statistical Analysis Help?
Our biostatisticians support you with sample size calculation to result interpretation. SPSS, R, and Stata expertise.
Get Stats Help →Our Statistics Services
📊 Thesis Statistics - ₹5,999 📈 Article Statistics - ₹4,999 📉 Meta-Analysis Stats - ₹14,999
Common Mistakes to Avoid
- Using wrong test: Parametric test on non-normal data
- Multiple testing: Running many tests increases false positives
- Ignoring assumptions: Each test has requirements
- Confusing correlation with causation
- Cherry-picking results: Report all findings, not just significant ones
- Inadequate sample size: Low power means unreliable results
Quick Reference for Your Viva
Be prepared to explain:
- Why you chose each statistical test
- Your sample size calculation
- What p-values mean in your context
- Confidence intervals interpretation
- Limitations of your statistical approach
The Bottom Line
You don't need to become a statistician. But you DO need to understand the basics of what tests you used and why. The actual calculations can be done by software or outsourced to experts.
What examiners and reviewers care about is whether you understand your own results. Focus on that, and your statistical analysis section will be solid.