You’ve done the heavy lifting: the Friedman test flagged that something’s off. The next step—file a post‑hoc comparison—turns that vague signal into actionable insight. Pick the right test, adjust for multiple comparisons, and report with clarity. Then you can confidently say, “Here’s exactly which treatment outperforms the others.” It’s not just data; it’s a story that guides decisions.

Which Of The Following Should Be Filed Immediately After Friedman: Complete Guide

Which of the Following Should Be Filed Immediately After Friedman?
The quick‑guide to the next step after the Friedman test.

Opening hook

You just ran a Friedman test and the p‑value is way below 0.File a post‑hoc test. It’s a common pause in data analysis: the test tells you something’s off, but it doesn’t tell you which groups differ. ” you ask. The answer? Your heart races. On the flip side, “What now? In practice, 05. And if you miss it, you’ll miss the real story behind the numbers.

What Is the Friedman Test?

The Friedman test is a non‑parametric alternative to the one‑way repeated measures ANOVA. That's why the data are paired, the normality assumption is shaky, and you still need a way to see if at least one method outperforms the others. Even so, imagine you’re comparing three teaching methods across the same class of students. That’s where Friedman steps in.

Key points

Paired data: Each subject is measured under every condition.
Ranks instead of raw scores: It ranks observations within each block (subject) and then compares those ranks across treatments.
Null hypothesis: All treatments have the same distribution of ranks.
Output: A chi‑square statistic and a p‑value.

In short, it’s a quick way to flag that something is different—but not which treatment is different.

Why It Matters / Why People Care

You’re not alone in hitting a Friedman test in your analysis. In real terms, many researchers, especially in psychology, education, and health sciences, rely on it for small sample sizes or ordinal data. But stopping at a significant Friedman result is like finding a red light on a road trip and deciding to keep driving without checking the map.

Misinterpretation risk: A significant result tells you “something’s off,” not “which pair is off.”
Decision‑making: If you’re choosing a new protocol or policy, you need to know which option is superior.
Reproducibility: Journals often require a post‑hoc test to pinpoint differences; otherwise, the paper feels incomplete.

So, what’s the next file you should produce? A post‑hoc test that digs deeper.

How It Works (or How to Do It)

After you run a Friedman test, the next logical step is a pairwise comparison that respects the repeated measures structure. Here’s the playbook.

### 1. Choose the Right Post‑Hoc Test

Test	When to Use	Key Feature
Dunn’s test (with Bonferroni correction)	Small to moderate sample sizes	Simple, controls family‑wise error
Conover’s test	Larger samples, more power	Uses exact distribution of ranks
Wilcoxon signed‑rank test (for each pair)	If you want the classic test	No adjustment for multiple comparisons unless you add it
Friedman + Nemenyi	When you want a global comparison	Less powerful but straightforward

### 2. Calculate the Test Statistic

Most statistical software (R, Python, SPSS, SAS) will give you the post‑hoc p‑values automatically once you specify the test. If you’re doing it by hand:

Rank the data within each block.
Compute the sum of ranks for each treatment.
Plug into the test’s formula (e.g., Conover’s uses the difference in rank sums).

### 3. Adjust for Multiple Comparisons

You’re making k(k‑1)/2 comparisons. If you have three treatments, that’s three pairwise tests. Use a correction:

Bonferroni: Divide alpha by the number of comparisons. Very conservative.
Holm‑Bonferroni: Step‑down procedure, less conservative.
False Discovery Rate (FDR): If you have many treatments, consider Benjamini–Hochberg.

### 4. Interpret the Results

Significant pair: That treatment pair differs in median rank.
Non‑significant pair: No evidence of a difference; may still be practical difference but not statistically proven.

### 5. Report Clearly

State the test used and the correction method.
Provide the test statistic, degrees of freedom, and p‑value for each pair.
Include a table or figure that visually shows the ranking differences.

Common Mistakes / What Most People Get Wrong

Skipping the post‑hoc test altogether
A significant Friedman result is only half the story. Without pairwise comparisons, you can’t tell which treatment is better.

Want to learn more? We recommend why planets don't fall into the sun and who is payee in cheque for further reading.
Using parametric post‑hoc tests
After a non‑parametric Friedman, you should stick to non‑parametric pairwise tests. Mixing parametric and non‑parametric can inflate Type I error.
Neglecting multiple‑comparison adjustments
Running raw Wilcoxon tests on every pair without correction is a recipe for false positives.
Reporting only the p‑value
Include effect sizes (e.g., rank-biserial correlation) to convey practical significance.
Assuming the same direction of effect across all pairs
The Friedman test only says something differs; the post‑hoc tells you which direction and by how much.

Practical Tips / What Actually Works

Use software that automates the whole process. In R, friedman.test() followed by pairwise.wilcox.test() with p.adjust.method="holm" is a solid workflow.
Create a visual aid early. A boxplot of ranks or a heatmap of pairwise differences helps readers grasp the pattern before diving into numbers.
Keep the sample size in mind. Non‑parametric tests are powerful with small samples, but post‑hoc tests lose power quickly. If you’re stuck with a tiny dataset, consider reporting descriptive statistics alongside the tests.
Document your correction method. Journals love transparency. Mention whether you used Bonferroni, Holm, or FDR.
Check assumptions. Even though Friedman is non‑parametric, the post‑hoc tests assume symmetry of the differences. If that’s violated, consider a bootstrap approach.

FAQ

Q1: Can I use a t‑test after a Friedman test?
No. The t‑test assumes normality and independence—both violated in repeated measures data. Stick to non‑parametric pairwise tests.

Q2: What if my post‑hoc test still shows no significant differences?
A significant Friedman test with non‑significant pairwise results suggests a global effect that isn’t driven by any single pair. You may need to explore higher‑order interactions or consider a different analytic approach.

Q3: Do I need to correct for multiple comparisons if I only have three treatments?
Yes, even with three treatments you’re doing three pairwise tests. A Bonferroni correction (α/3) is common, but you can use Holm‑Bonferroni for a less conservative approach.

Q4: My software didn’t give me a post‑hoc option. What do I do?
Manually run pairwise Wilcoxon signed‑rank tests and apply a correction. Most statistical packages allow you to specify the correction method.

Q5: Is it okay to report the Friedman test result alone in a paper?
Journals typically expect a post‑hoc analysis to identify which groups differ. If you can’t run one, explain why and discuss the limitations.

Closing paragraph

You’ve done the heavy lifting: the Friedman test flagged that something’s off. Pick the right test, adjust for multiple comparisons, and report with clarity. Worth adding: then you can confidently say, “Here’s exactly which treatment outperforms the others. The next step—file a post‑hoc comparison—turns that vague signal into actionable insight. ” It’s not just data; it’s a story that guides decisions.