Visualizing Data Relationships with Seaborn Pair Plots

Making a Seaborn pair plot is ridiculously easy. I mean, you know, assuming you already know how to do it. It's a very short amount of code.

So, you know, maybe 'easy' isn't the right word. It's straightforward. SNS, which is typically how we import Seaborn.

Which, by the way, doesn't stand for anything. Nobody knows why we started doing it as SNS. It's lost to the mists of time.

Seaborn users today—we don't know. We say, 'Make a pair plot from our car sales.' And it will take a look, just like we did when we asked for a correlation matrix.

It can do the correlation matrix and make a plot of plots. Then we say, hey, pi plot, show that. And it will fail to auto—oh, there it is.

And it'll take a moment. It takes a second to make a pair plot. It's 25 graphs.

There they are. Let's take a look at these. Ooh, they're slightly cut off.

Ah, we need to make this a little bit bigger at the top. No, there we go. It was just, it was just scrolled.

---

We could, you know, resize this, but it seems like it's totally fine. Don't panic, everybody. Stop panicking.

Everything looks great. Okay, so what is this? Again, when you are looking at the relationship between sales in thousands and sales in thousands, right? Sales in thousands at the top, sales in thousands down here. Yeah—I'm going to zoom out.

Yeah, sales in thousands to sales in thousands. All down this line, we are comparing the thing to itself. So rather than show 1.0, it's showing us a histogram.

What we looked at earlier, when you're looking at distributions. So this is the distribution of sales in thousands. A lot of items, you know, selling at this lower price point and just like one thing selling way over here.

And the same goes for fuel efficiency. Here's the distribution of fuel efficiency. You can see these are vague bell curves.

There's not a lot of data, so they're not smooth, but they are roughly normal distributions—price in thousands, engine size, and so on. But the interesting thing here is each one's relationship to other things. A lot of them are blobs.

'Sales in thousands' doesn't really seem to correlate with anything. Again, we can really see from this graph why our domain knowledge was wrong. The data shows that there's no relationship with anything here.

---

Now, other pairs have a stronger relationship. The most obviously strong one here is horsepower to price in thousands. That seems like a really good predictor overall.

Horsepower versus price in thousands. Now, there are other ones that are strong predictors, some of them in the other direction, right? If we look at fuel efficiency and engine size, engine size, fuel efficiency, right? Engine size here on the left, sorry, fuel efficiency here on the left, engine size here on the bottom. They're negatively correlated.

They're going down and to the right. So there's a strong correlation there. It's just in the opposite direction. The higher the engine size, the lower the fuel efficiency.

But let's take a look at price in thousands. That's what we're most interested in. Again, no real relationship here.

Some relationship to fuel efficiency. There's a down and to the right relationship here. The higher the fuel efficiency, the lower the price in thousands.

And then horsepower and engine size have fairly strong, particularly horsepower relationships here. So this is a great way to really visualize this, not just as numbers, but as different data points and seeing their relationship. Pair plots—one of the best visualization tools.