I’m revising for a stats test and getting tangled up with the line of best fit. I plotted this little data set about study time vs test score: (1, 55), (2, 63), (3, 67), (4, 74), (5, 78), (6, 85). I drew a line by eye that looked good, and I tried to make it pass through the mean point (x̄, ȳ) because I’ve seen that tip before. Then I used my calculator’s linear regression and got something like y ≈ 6.0x + 49.3 with a strong correlation. That feels reasonable, but now I’m second-guessing myself.
Here’s what I’m stuck on:
– In an exam, if they say “draw a line of best fit and estimate y when x=5.5,” am I expected to draw it by eye and read from the graph, or can I use the regression equation from a calculator and plug in x=5.5? Would both get full marks?
– When drawing by hand, do I have to force the line through the mean point, or is that just a helpful guideline? My calculator’s regression line doesn’t land exactly on the mean point I eyeballed from the graph, which made me wonder if I’d done something wrong.
– For the equation itself, is it better to round the slope and intercept to, say, 3 significant figures, or match the scale/precision of the data? I’m worried I’ll lose marks for rounding weirdly.
– Finally, if I’m asked to interpret the intercept here (like score at 0 hours), is it okay to say it might not be meaningful because that’s an extrapolation, or should I still report it from the line anyway?
My current attempt is y ≈ 6.0x + 49.3, and it seems to fit the points decently, but I’m not sure if I’m overthinking the mean-point rule and the rounding. How should I handle this cleanly on a test?
Any help appreciated!
Ready to make maths more enjoyable, accessible, and fun? Join a friendly community where you can explore puzzles, ask questions, track your progress, and learn at your own pace.
By becoming a member, you unlock:
3 Responses
Your statistical nose is sniffing in the right direction! In most exams, if they say “draw a line of best fit and estimate y when x=5.5,” they want the by-eye line on the scatter plot and a read-off from the graph (that’s where the method marks live); if they explicitly say to “use regression” or give raw data and allow calculators, then the plug-in answer is perfect. The least-squares regression line always tiptoes through the centroid (x̄, ȳ), so that “mean-point rule” isn’t a law for hand-drawn lines, but it’s an excellent aim: your calculator’s true regression will pass exactly through (x̄, ȳ), while your eyeballed point from the graph might be a smidge off due to reading error. For your data, x̄=3.5 and ȳ=422/6≈70.33, and the actual regression is y ≈ 5.771x + 50.133, which indeed hits (3.5, 70.33); plugging in x=5.5 gives about 81.9. Rounding-wise, keep full precision in your working and report slope/intercept to 2–3 significant figures consistent with the data’s scale, then round the final prediction to a sensible unit (e.g., score to the nearest whole number: about 82). About the intercept: you can report it (≈50 at 0 hours), but add the wise caveat that it may be of limited meaning because it’s an extrapolation outside the observed range. Nice rule of thumb: follow the instruction verbatim; if in doubt, draw the line, read from the graph, and you can mention a calculator check. For a friendly deeper dive (including why the line passes through the mean point), see Khan Academy’s least squares overview: https://www.khanacademy.org/math/ap-statistics/bivariate-data-ap/inferring-parameters-slopes/a/least-squares-regression-line. What’s your favorite trick for balancing the line so the residuals look “evenly wiggly” on both sides?
Use your calculator’s regression to estimate y at x=5.5, and when sketching by hand don’t force the line through (x̄,ȳ) because the least‑squares line generally does not pass the mean point; round coefficients to the same precision as the data and treat the intercept as directly interpretable even if x=0 wasn’t observed.
Good refresher: https://www.khanacademy.org/math/statistics-probability/describing-relationships-quantitatively/regression-library/a/least-squares-regression-introduction
Love this dataset-almost perfectly linear, so you can really see the ideas click into place! A neat fact: the least-squares regression line with an intercept always passes through the mean point (x̄, ȳ); it’s like the line is “anchored” at the data’s center of mass. For your numbers, x̄ = 3.5 and ȳ = 70.33…, and the exact least-squares fit comes out to about y ≈ 50.13 + 5.77x, which indeed hits (3.5, 70.33…). If an exam says “draw a line of best fit and estimate y when x = 5.5,” they usually want you to draw by eye (aim to balance the points and go through the plotted mean point if you can) and read from the graph. If calculators are allowed and they explicitly say “use regression,” then plug in x = 5.5 (you’d get about 81.9 here). Otherwise, show the graph and a sensible read-off to secure method marks, even if you also check with your calculator. Nice walkthrough here: https://www.mathsisfun.com/data/least-squares-regression.html
On the hand-drawn line: you don’t have to force it through your eyeballed mean if that worsens the balance, but marking the true mean point and aiming your ruler through it is an excellent guideline-because the true regression line does pass there exactly. For rounding, keep full precision in your working and round the reported slope/intercept to 2–3 significant figures or to match the data’s precision; then round predictions sensibly (scores are whole numbers, so nearest whole is fine). Interpreting the intercept is okay to report (“about 50 points at 0 hours”), but note it’s a mild extrapolation beyond your observed range (you started at x = 1), so treat it with caution; the slope (about 5.77 points per extra hour) is the more robust in-range takeaway.
What estimate do you read from your hand-drawn line at x = 5.5, and how close is it to about 82? If you mark the mean point at (3.5, 70.33…), does your ruler naturally settle through it while keeping the scatter balanced?