When we want to test construct validity, we can often start with correlation. This tells us how strongly two variables move together. Let’s say we have a new questionnaire to measure fatigue. To check its validity, we might compare it with an existing, trusted fatigue scale. If both tools measure the same thing, their scores should be highly correlated. Get an understanding of construct validity.
The most common correlation measure is Pearson’s r. It’s calculated like this:
Where
Xi and Yi are individual scores, X̄ and Ȳ are the means. The top part shows shared variation. The bottom adjusts for how spread out the values are.
In simple terms, it’s a ratio of shared variation to total variation.
Why Correlation?
If two measures claim to capture the same construct (like depression), their scores should rise and fall together. That’s convergent validity. A strong positive correlation (say, r=0.7r = 0.7r=0.7 or higher) supports this. If the new scale measures something different, say, physical strength, we expect little to no correlation. That’s discriminant validity. A low or near-zero r supports that the constructs are unrelated.
Let us look into an Example
Imagine:
- Patient A scores 8 on the new fatigue scale and 9 on the old one.
- Patient B scores 2 and 3.
- Patient C scores 6 and 6.
We plot these pairs and calculate Pearson’s r. If it comes out high, the new scale aligns well with the old one. That gives us evidence that it truly measures fatigue.
Correlation doesn’t prove causation. And a high r doesn’t mean two tools are identical. But in construct validation, it’s a powerful first check.