Comparing distributions

Think of it like being a detective trying to compare two different groups of things, like two piles of toys or two different classes' test scores. You want to know if one pile is bigger, if the toys in one pile are older, or if one class generally did better than the other.

In Statistics, when we talk about "comparing distributions," we're looking at how data (which is just a fancy word for information or numbers) is spread out or arranged for two or more different groups. We want to see if these groups are alike or different in important ways.

We usually compare them using four main features, which you can remember with the acronym C.U.S.S.:

• Center: Where is the 'middle' or 'typical' value for each group? (Like the average height of kids in two different schools).
• Unusual features: Are there any weird points, like outliers (numbers that are much bigger or smaller than the rest), or gaps in the data?
• Shape: What does the 'picture' of the data look like? Is it lopsided, symmetrical, or does it have multiple peaks?
• Spread: How much do the numbers in each group vary? Are they all really close together, or are they scattered far apart? (Like if one class's test scores were all 80s, and another class had scores from 20s to 100s).

By comparing these four things, we can paint a clear picture of how our groups are similar and different.

Let's say a video game company wants to compare the playtime (how long people play) of two new games, 'Game A' and 'Game B', to see which one is more engaging. They collect data from 100 players for each game.

1. Look at the Center: They might find that the median (the middle value when all playtimes are lined up) playtime for Game A is 3 hours, while for Game B it's 5 hours. This tells them that, on average, people play Game B longer.
2. Look at Unusual Features: For Game A, they might see one player who played for 20 hours – that's an outlier! Maybe that player is a super fan or a tester. For Game B, all playtimes might be pretty close together, with no unusual long or short sessions.
3. Look at the Shape: When they make a histogram (a bar graph showing how often different playtimes occur) for Game A, it might be skewed right (meaning most people play for short times, but a few play for very long times, pulling the 'tail' of the graph to the right). Game B's histogram might be more symmetrical (like a bell curve), meaning playtimes are evenly spread around the middle.
4. Look at the Spread: They might notice that Game A's playtimes range from 1 hour to 20 hours (a big range), meaning players have very different engagement levels. Game B's playtimes might only range from 3 hours to 7 hours, meaning most players have similar engagement. This means Game B has less variability (less spread).

By comparing these C.U.S.S. features, the company learns that Game B generally keeps players engaged longer and more consistently, even though Game A has a few super-dedicated players.

When you're asked to compare two or more distributions, follow these steps:

1. Visualize the Data: First, create appropriate graphs for each group, like dot plots, histograms, or box plots. This helps you 'see' the data.
2. Identify the Center: Find a measure of the middle for each group, usually the mean (average) or median (middle value).
3. Note Unusual Features: Look for any outliers (data points far from the rest) or gaps in the data for each group.
4. Describe the Shape: For each graph, describe if it's symmetrical, skewed left (tail to the left), skewed right (tail to the right), or has multiple peaks (modes).
5. Measure the Spread: Determine how spread out the data is for each group, using range, interquartile range (IQR), or standard deviation.
6. Compare and Contrast: Finally, write a paragraph (or two!) comparing all these features side-by-side for each group. Use comparison words like "greater than," "less than," "similar to," or "more variable than."