|On this page…|
Normal probability plots are used to assess whether data comes from a normal distribution. Many statistical procedures make the assumption that an underlying distribution is normal, so normal probability plots can provide some assurance that the assumption is justified, or else provide a warning of problems with the assumption. An analysis of normality typically combines normal probability plots with hypothesis tests for normality.
This example generates a data sample of 25 random numbers from a normal distribution with mu = 10 and sigma = 1, and creates a normal probability plot of the data.
rng('default'); % For reproducibility x = normrnd(10,1,25,1); normplot(x)
The plus signs plot the empirical probability versus the data value for each point in the data. A solid line connects the 25th and 75th percentiles in the data, and a dashed line extends it to the ends of the data. The y-axis values are probabilities from zero to one, but the scale is not linear. The distance between tick marks on the y-axis matches the distance between the quantiles of a normal distribution. The quantiles are close together near the median (probability = 0.5) and stretch out symmetrically as you move away from the median.
In a normal probability plot, if all the data points fall near the line, an assumption of normality is reasonable. Otherwise, the points will curve away from the line, and an assumption of normality is not justified. For example, the following generates a data sample of 100 random numbers from an exponential distribution with mu = 10, and creates a normal probability plot of the data.
x = exprnd(10,100,1); normplot(x)
The plot is strong evidence that the underlying distribution is not normal.
Quantile-quantile plots are used to determine whether two samples come from the same distribution family. They are scatter plots of quantiles computed from each sample, with a line drawn between the first and third quartiles. If the data falls near the line, it is reasonable to assume that the two samples come from the same distribution. The method is robust with respect to changes in the location and scale of either distribution.
To create a quantile-quantile plot, use the qqplot function.
The following example generates two data samples containing random numbers from Poisson distributions with different parameter values, and creates a quantile-quantile plot. The data in x is from a Poisson distribution with lambda = 10, and the data in y is from a Poisson distribution with lambda = 5.
x = poissrnd(10,50,1); y = poissrnd(5,100,1); qqplot(x,y);
Even though the parameters and sample sizes are different, the approximate linear relationship suggests that the two samples may come from the same distribution family. As with normal probability plots, hypothesis tests can provide additional justification for such an assumption. For statistical procedures that depend on the two samples coming from the same distribution, however, a linear quantile-quantile plot is often sufficient.
The following example shows what happens when the underlying distributions are not the same. Here, x contains 100 random numbers generated from a normal distribution with mu = 5 and sigma = 1, while y contains 100 random numbers generated from a Weibull distribution with A = 2 and B = 0.5.
x = normrnd(5,1,100,1); y = wblrnd(2,0.5,100,1); qqplot(x,y);
These samples clearly are not from the same distribution family.
An empirical cumulative distribution function (cdf) plot shows the proportion of data less than each x value, as a function of x. The scale on the y-axis is linear; in particular, it is not scaled to any particular distribution. Empirical cdf plots are used to compare data cdfs to cdfs for particular distributions.
The following example compares the empirical cdf for a sample from an extreme value distribution with a plot of the cdf for the sampling distribution. In practice, the sampling distribution would be unknown, and would be chosen to match the empirical cdf.
y = evrnd(0,3,100,1); cdfplot(y) hold on x = -20:0.1:10; f = evcdf(x,0,3); plot(x,f,'m') legend('Empirical','Theoretical','Location','NW')
A probability plot, like the normal probability plot, is just an empirical cdf plot scaled to a particular distribution. The y-axis values are probabilities from zero to one, but the scale is not linear. The distance between tick marks is the distance between quantiles of the distribution. In the plot, a line is drawn between the first and third quartiles in the data. If the data falls near the line, it is reasonable to choose the distribution as a model for the data.
To create probability plots for different distributions, use the probplot function.
The following example assesses two samples, one from a Weibull distribution with A = 3 and B = 3, and one from a Rayleigh distribution with B = 3, to see if either distribution may have come from a Weibull population.
x1 = wblrnd(3,3,100,1); x2 = raylrnd(3,100,1); probplot('weibull',[x1 x2]) legend('Weibull Sample','Rayleigh Sample','Location','NW')
The plot gives justification for modeling the first sample with a Weibull distribution; much less so for the second sample.
A distribution analysis typically combines probability plots with hypothesis tests for a particular distribution.