A Q-Q (quantile-quantile) plot is a scatterplot that compares your sample data against a theoretical normal distribution to visually assess how closely they match. You enter a column of numbers and the calculator produces the plot’s underlying quantile pairs; the closer the points lie to the diagonal reference line, the more your data resemble a normal distribution. This page explains how to read the plot, common pitfalls, and what to do when points deviate.
Formula
The Q-Q plot pairs each sorted sample value with a theoretical quantile from the standard normal distribution. For the i-th smallest observation out of n total, the theoretical quantile is:
zi = Φ−1( (i − 0.5) / n )
where Φ−1 is the inverse standard normal CDF (the z-score), i is the rank (1 for the smallest), and n is the sample size. The subtraction of 0.5 is a continuity correction that places each point at the middle of its probability interval. The plot then shows the pairs ( zi , sorted data value ). If the data are normally distributed, the points fall roughly on the line y = mean + SD × z.
Worked Example
Using the default data: 1.2, 2.3, 0.9, 3.1, 1.8, 2.7, 1.5, 2.0, 0.6, 1.4. First sort the data: 0.6, 0.9, 1.2, 1.4, 1.5, 1.8, 2.0, 2.3, 2.7, 3.1. For n = 10, the theoretical quantiles are:
- i=1: Φ−1(0.05) = −1.645
- i=2: Φ−1(0.15) = −1.036
- i=3: Φ−1(0.25) = −0.674
- i=4: Φ−1(0.35) = −0.385
- i=5: Φ−1(0.45) = −0.126
- i=6: Φ−1(0.55) = 0.126
- i=7: Φ−1(0.65) = 0.385
- i=8: Φ−1(0.75) = 0.674
- i=9: Φ−1(0.85) = 1.036
- i=10: Φ−1(0.95) = 1.645
Plotting these against the sorted data gives a cloud of points that should be roughly linear. The calculator outputs the same pairs. For this small sample, points lie close to a straight line, suggesting no strong departure from normality.
Assumptions
- The data are a random sample from the population of interest.
- Observations are independent of each other.
- The theoretical distribution is the standard normal (mean 0, SD 1) for the quantiles; the sample line can have any slope and intercept.
How to Read a Q-Q Plot
The reference line is not fixed at 45°; it is the line through the first and third quartiles of the data, or equivalently the line y = mean + SD·z. Points that follow this line indicate normality. Deviations reveal specific features:
- Both ends curve up (S-shape): heavy tails – more extreme values than normal.
- Both ends flatten (inverse S): light tails – fewer outliers.
- Points curve above the line on the right: right-skew (positive skew).
- Points curve below on the left: left-skew (negative skew).
- Gaps or jumps: possible outliers or a mixture of two populations.
In practice, small samples (n < 30) can show random wiggles even when data are truly normal. Do not over-interpret minor bends; combine the plot with a formal test like Shapiro–Wilk if you need a decision.
When to Use a Q-Q Plot
Use a Q-Q plot when you need to check the normality assumption before running parametric tests (t-tests, ANOVA, linear regression). It is more reliable than a histogram because it does not depend on bin width. You also use it to compare data against other distributions (exponential, uniform) by changing the theoretical quantiles, but the normal version is the most common. The plot works best with at least 20–30 points; with fewer, the quantile estimates are too noisy.
Common Mistakes and Misconceptions
- Expecting the points to lie on a 45° line. The reference line is the line of best fit through the quartiles, not a fixed diagonal.
- Ignoring sample size. A small sample can look non-normal by chance; a large sample can show trivial deviations that are statistically significant but practically irrelevant.
- Using the plot alone for very small n. For n < 10, the plot is nearly useless; use a formal test or collect more data.
- Confusing skew with outliers. A single extreme outlier often appears as an isolated point far from the line, while skew bends a whole tail.
- Forgetting that the plot is not a hypothesis test. It is a visual aid; it does not give a p-value. Pair it with Shapiro–Wilk or Anderson–Darling for a quantitative answer.
Alternatives to the Q-Q Plot
If you need a numeric check, the Shapiro–Wilk test is the most powerful for normality, followed by Anderson–Darling and Kolmogorov–Smirnov (with Lilliefors correction). For visual checks, a histogram with an overlaid normal curve or a boxplot can reveal skew and outliers, but both lose information compared to a Q-Q plot. In regression, you can plot residuals against fitted values to check homoscedasticity, and a Q-Q plot of residuals to check normality. For very large datasets (n > 10,000), the Q-Q plot will almost always show some deviation because real data are rarely perfectly normal; focus on whether the deviation affects your analysis.
Frequently Asked Questions
What does a Q-Q plot tell me? It tells you whether your sample’s quantiles match the quantiles of a normal distribution. If the points follow a straight line, the data are approximately normal; systematic curvature indicates skew or heavy tails.
How do I know if my Q-Q plot is good enough? There is no strict rule. As a guideline, if the points stay within the confidence band (if drawn) or if the deviations are small relative to the slope, normality is plausible. For n < 30, accept minor wiggles; for n > 200, even small bends may be real but often harmless.
Can I use a Q-Q plot for non-normal distributions? Yes, you can replace the normal quantiles with quantiles from any distribution (e.g., exponential, log-normal). The interpretation is the same: points on a line mean the data fit that distribution.
Why do my points curve at the ends? Curving at both ends typically means your data have heavier tails (more extreme values) than normal, or lighter tails if the curve flattens. This is common in financial returns (heavy tails) or bounded measurements (light tails).
Is a Q-Q plot better than a histogram? For normality checks, yes. A histogram’s appearance changes with bin width, while a Q-Q plot is bin-independent and shows deviations more clearly, especially in the tails.
What should I do if my data are not normal? Options include transforming the data (log, square root, Box–Cox), using non-parametric tests (Wilcoxon, Kruskal–Wallis), or using robust methods that do not assume normality. The choice depends on your analysis goal and sample size.
Follow your lab method, the safety data sheet (SDS) and your supervisor when performing any bench work.
Related
Formula used
in list | Data values | 1.2, 2.3, 0.9, 3.1, 1.8, 2.7, 1.5, 2.0, 0.6, 1.4 | 0 out theo_q | Theoretical quantiles (z-scores) | - | qnorm((rank(data,1)-0.5)/count(data)) | 3 out sample_q | Sample quantiles (sorted data) | - | sort(data) | 3
Each line is one input (in, sel, list) or one result (out: label, unit, formula, decimals). How the calculators work
Method: the standard formula for this calculation with the constants and conversion factors written on the page, rounded as each result states. See method and constants, the disclaimer or report an error.
More in this section
- Normality to g/L Converter & Formula29.220 g/L
- Normality to meq/mL Conversion Calculator1.0000 meq/mL
- Normality Test Q-Q Plot: How to Read It
- Normality Test Results: How to Analyze Them
- Normality Slay the Spire?
- Normality Test for Paired t Test: How to Check