Marks against hours studied

New word — scatter plt.scatter(df["hours"], df["marks"]). Two columns of the same length, taken pairwise: the first student's hours with the first student's marks, and so on down the file.
The order of the rows does not matter. Shuffle the file and the picture is identical, because each dot carries both of its numbers with it.

Careful — scatter, not plot plt.plot() joins the points with a line in file order. On data that arrives sorted, that line looks meaningful and is not — it is a join-the-dots of whatever order the rows happened to be in. Use plot for something measured over time, where the order is real. Use scatter when you are comparing two measurements of the same thing.

What to look for Direction — does the cloud rise to the right, fall, or lie flat?
Strength — is it a tight band or a shapeless blob? A tight rising band means hours tell you a lot about marks; a blob means they tell you almost nothing.
Oddities — a dot far from the rest is worth asking about, exactly as in problem 24.

New word — line of best fit np.polyfit(x, y, 1) works out the straight line that passes closest to all the points, and hands back two numbers: the slope and the intercept. The line itself is then m * x + c — the same y = mx + c you know from school.
A slope of about 4 means that each extra hour of study goes with about four more marks. It does not mean the hour caused them.

Two files, and the difference is the point study_strong.csv — the same ten students, where hours and marks rise together.
study_none.csv — the same hours, with marks that have nothing to do with them.
Both files fit a line, because a line can always be fitted. On the second one the line comes out almost flat and the dots are scattered nowhere near it. Fitting a line never tells you whether fitting a line was worth doing — your eyes do that, and problem 28 puts a number on it.