Comparing Distributions with Different Means

A distribution has two separate stories to tell: where its values are centered, and how spread out they are around that center. Two classes can have the exact same standard deviation but sit far apart on the number line, because standard deviation only measures how much values vary from their own mean, not where that mean is. A test where everyone scores near 60 and a test where everyone scores near 90 can have identical spread if the pattern of variation is the same shape, just relocated.

This is why shifting matters as a mental model. If you take every value in a data set and add the same number k to it, the whole set slides over by k. The mean moves by k, since the mean is just a balancing point and the balance point slides with everything else. But the standard deviation does not change, because standard deviation depends on the distances between values and between each value and the mean, and a shift preserves every one of those distances. Add 10 to every score in a class and the mean goes up by 10, but how spread out the scores are stays exactly the same.

This separation of center and spread connects directly to reading scatterplots and summary statistics elsewhere in data analysis: a trend line or a reported mean only tells you about typical behavior, not about every individual case. Knowing that a shift changes one summary statistic and not the other also helps you reason quickly about transformed data sets without recalculating everything from scratch.