When we test something with statistics, it’s like playing a guessing game. First, we set a rule (called α, usually 0.05) to decide if what we see is just luck or really true. Then we do the test and get some numbers (the p-value and test statistic). After that, we check the effect size – this tells us if the difference is small like a drop of water, or big like a bucket. Finally, we explain the result in everyday words, like: “The new toy really made kids happier, and not just by chance, but in a way that really matters.”
| The girl thinks, “I’ll only believe it if it really works.” Being 95% sure means we trust the result is real but accept a 5% risk of being wrong-this is significance. It’s like saying, “I’m sure it would work.” But how much it worked (clearing acne and oil) is shown in the right image-that’s the effect size. | After a week, she notices her acne is clearer and her skin less oily. This shows the effect size-not just whether the change is real, but how big and meaningful the change actually is in her life. |
How Inferential Statistics is Done – Step by Step
We will discuss the topics in red–
- Identify type of data (categorical / numerical, scale of measurement).
- Check assumptions (normality, variance, parametric vs non-parametric).
- Select appropriate statistical test (based on data type, groups, design).
- Set significance level (α, usually 0.05).
- Run the test (get test statistic & p-value).
- Check effect size (strength of relationship/difference).
- Interpret results in plain language.
- Report findings (test, statistic, p-value, effect size).
Set significance level (α, usually 0.05)
Setting a significance level (α) is like deciding your tolerance for being wrong before you start. Usually, we pick 0.05, which means: “I am okay with being wrong 5 times out of 100. If the chance of my result being just luck is less than 5%, I’ll believe it’s real.” It’s basically the rule of the game in statistics. For example, imagine a student learns a new topic and then appears for a test. The teacher thinks: “If the student scores well, I’ll believe they truly understood the topic-but only if there’s less than a 5% chance that the good score happened by pure luck (like guessing).” That 5% cut-off (α = 0.05) means the teacher is actually 95% confident.
In statistics, we usually keep the significance level at 5% (α = 0.05).
This means we are 95% confident that our result is real and not just luck.
Example: A student gives a test. If the chance of their high score being just guesswork is less than 5%, the teacher accepts that the student really understood the topic.Memory trick:“5% risk, 95% trust.”
Why 5%? It’s not a magic number-it’s just a balance. Too loose (like 20%) and we might give credit even when the student guessed. Too strict (like 1%) and we might doubt even when the student really learned. So, 5% is the fair middle ground most researchers use.
Run the test (get test statistic & p-value)
Once you’ve chosen the right statistical test during the “select appropriate test” step (based on your data type, number of groups, and design), now is the time to apply it actually. Running the test gives you two main numbers:
- the test statistic → a number that shows how big the difference or relationship is compared to what we’d expect by chance.
- the p-value → the probability that your result happened just by luck.
- Blog on p-value for better understanding
| Example: Imagine the teacher now checks the student’s test answers using the chosen method (like checking multiple-choice with an answer key). The score (test statistic) tells how well the student did, and the chance of guessing correctly (p-value) tells |
Check effect size (strength of relationship/difference)
Finding significance tells us “Is the result real, or just luck?” But it doesn’t tell us “How big or important is the result?” That’s where effect size comes in. It measures the strength of the difference or relationship.Example: Imagine a student passes the test. Significance tells the teacher: “Yes, the student really learned.” But effect size tells: “Did the student just barely pass cuz they understood some parts of it, or did they score so high that it’s clear they mastered the topic?”
In short, significance = Is it real?
Effect size = How strong or meaningful is it?
Interpret results in plain language
| He fed his data into the computer, but all he got back were numbers (p = 0.01, d = 0.8). Now he’s frustrated, wondering what these numbers actually mean in plain English. |
After checking significance and effect size, the final step is to translate numbers into a clear story. Instead of just saying “p < 0.05, Cohen’s d = 0.8,” we explain what that really means in everyday words.Example: A teacher wouldn’t tell parents, “The test statistic was 3.5 with p = 0.01.” Instead, they’d say: “Your child’s new study method clearly worked, and the improvement was strong, not just by chance.”
Report findings (test, statistic, p-value, effect size).
The last step is to put everything together so others can see exactly what you did and why your result matters. Reporting is important because it makes your work transparent, clear, and easy to understand-not just for statisticians, but for teachers, doctors, or anyone reading your research.
A proper report includes:
- Research Question / Hypothesis
- Variables & Data Type
- Study Design
- Sample Details
- Assumption Checks
- Significance Level (α)
- Chosen Statistical Test
- Run the Test
- Effect Size
- Interpretation in Plain Language
- Report Findings Clearly
Choosing the right statistical test and checking assumptions like normality and equal variances ensures reliable results. Parametric tests work well for normal data, while non-parametric tests help with non-normal or categorical data. This careful approach makes research conclusions trustworthy.
10 CUET-style MCQs
1. What does a significance level (α = 0.05) indicate?
A) The probability of rejecting a true null hypothesis (Type I error)
B) The strength of the relationship
C) The average difference between groups
D) The minimum sample size required
2. Which of the following is NOT part of inferential statistics?
A) Test statistic
B) p-value
C) Effect size
D) Median
3. Effect size tells us:
A) Whether the result is significant
B) How strong or meaningful the effect is
C) The sample size needed
D) The confidence interval
4. Non-parametric tests are typically used when:
A) Data is normally distributed
B) Data is categorical or non-normal
C) Sample size is very large
D) Variables are measured at interval/ratio level
5. A p-value represents:
A) The probability of obtaining results as extreme as observed if the null is true
B) The size of the effect
C) The pre-decided significance level
D) The type of test chosen
6. After selecting a statistical test, the researcher should first:
A) Check assumptions
B) Report findings immediately
C) Interpret results
D) Set a new hypothesis
7. Why is α = 0.05 commonly chosen?
A) It is considered a conventional balance between confidence and risk
B) It guarantees accurate conclusions
C) It depends on the sample size
D) It ensures a large effect size
8. Which is the correct sequence of steps in inferential analysis?
A) Select test → Interpret → Report → Set α → Check assumptions
B) Report → Run test → Select test → Interpret → Set α
C) Check assumptions → Set α → Select test → Run test → Interpret → Report
D) Interpret → Check assumptions → Report → Run test → Set α
9. Interpreting results in plain language is important because:
A) It makes findings accessible to non-experts
B) It replaces hypothesis testing
C) It avoids the need for p-values
D) Only statisticians can understand numbers otherwise
10. Which statement about parametric tests is correct?
A) They are always better than non-parametric tests
B) They are only used for categorical data
C) They require assumptions such as normality and equal variances
D) They ignore sample size
Mapping answers
1 – A
2 – D
3 – B
4 – B
5 – A
6 – A
7 – A
8 – C
9 – A
10 – C
Leave a Reply