Mathwords logoMathwords

P-P Plot — Definition, Formula & Example

A P-P plot is a graph that checks whether data follow a proposed probability distribution. It plots the proportion of the sample at or below each data value against the probability the model gives to values at or below it. If the model fits, the points lie close to the diagonal line y=xy = x.

Let x(1)≤x(2)≤⋯≤x(n)x_{(1)} \le x_{(2)} \le \dots \le x_{(n)} be the ordered sample, and let FF be a fully specified hypothesized cumulative distribution function (CDF). The P-P plot is the set of points (F(x(i)), pi)\big(F(x_{(i)}),\, p_i\big) for i=1,…,ni = 1, \dots, n. Here pip_i is an empirical plotting position, such as in+1\tfrac{i}{n+1} or i−0.5n\tfrac{i - 0.5}{n}. Both coordinates lie in [0,1][0, 1], and the reference line is y=xy = x.

Key Formula

(F(x(i)),  in+1),i=1,2,…,n\left(F\big(x_{(i)}\big),\; \frac{i}{n+1}\right), \quad i = 1, 2, \dots, n
Where:
  • x(i)x_{(i)} = The i-th smallest data value
  • FF = CDF of the hypothesized distribution, with all parameters specified
  • nn = Sample size
  • in+1\frac{i}{n+1} = Empirical plotting position, one common choice among several

Worked Example

Problem: A dart player claims her darts land evenly over the area of a circular target with radius 10 inches. If that is true, the distance RR from the center has CDF F(r)=πr2π⋅102=r2100F(r) = \frac{\pi r^2}{\pi \cdot 10^2} = \frac{r^2}{100} for 0≤r≤100 \le r \le 10. Four throws land 4, 6, 8, and 9 inches from the center. Build a P-P plot and judge the fit.
Step 1: The data are already sorted, and n=4n = 4. Assign the plotting positions pi=in+1=i5p_i = \frac{i}{n+1} = \frac{i}{5}.
p1=0.2,p2=0.4,p3=0.6,p4=0.8p_1 = 0.2,\quad p_2 = 0.4,\quad p_3 = 0.6,\quad p_4 = 0.8
Step 2: Evaluate the model CDF at each ordered value.
F(4)=16100=0.16,F(6)=0.36,F(8)=0.64,F(9)=0.81F(4) = \frac{16}{100} = 0.16,\quad F(6) = 0.36,\quad F(8) = 0.64,\quad F(9) = 0.81
Step 3: Plot the pairs (F(x(i)),pi)(F(x_{(i)}), p_i) and compare each point with the line y=xy = x. The vertical gaps are small.
∣0.2−0.16∣=0.04,  ∣0.4−0.36∣=0.04,  ∣0.6−0.64∣=0.04,  ∣0.8−0.81∣=0.01|0.2 - 0.16| = 0.04,\; |0.4 - 0.36| = 0.04,\; |0.6 - 0.64| = 0.04,\; |0.8 - 0.81| = 0.01
Answer: The points are (0.16, 0.2), (0.36, 0.4), (0.64, 0.6), and (0.81, 0.8). Each one lies within 0.04 of the line y = x, so the data are consistent with the claimed model. With only four throws, this is weak evidence, and a larger sample would give a more reliable check.

Why It Matters

Regression and ANOVA courses use P-P plots to check whether residuals are approximately normal before trusting p-values (with parameters estimated from the data, the plot is approximate). Reliability engineers use them to test whether failure times follow a proposed lifetime model. The plot shows where the model fails, which a single test statistic cannot do.

Common Mistakes

Mistake: Treating a P-P plot and a Q-Q plot as interchangeable.
Correction: A P-P plot compares probabilities and is most sensitive to misfit near the center of the distribution. A Q-Q plot compares quantiles, measured in data units, and shows tail problems more clearly. If you are worried about outliers or heavy tails, use a Q-Q plot.

Related Terms