Pearson System of Distributions

pearson family of distributions
The Pearson system of distributions.

The Pearson system of distributions, named after Karl Pearson (1857 to 1936), includes a wide range of curves including:

Differential Equations in Pearson System of Distributions

All curves in the system satisfy a differential equation of the form [1]

pearson family differential equation

The exact shape of the distribution depends on the values of parameters a and c.

If –a is not a root of the denominator set to equal zero, i.e.,

pearson family denominator

then p is finite when x = –a and dp/dx = 0. When p = 0, the slope is zero. If x ≠ –a and p # 0, then dp(x)/dx ≠ 0. However, two conditions must be satisfied:

  1. P(x) 0: The probability of any “x” must be greater than or equal to zero, and
  2. Like all probability density functions, the area under the curve must equal 1. In other words, it satisfies the integral
pearson family of distributions integral

The differential equation

The differential equation tells us that p(x) must tend to zero as x tends to infinity, so must dp/dx.  In some formal solutions, the condition P(x) ≥ 0 is not satisfied, so we must restrict the range of x to values which meet the condition. When x is outside the specified range, we can assign the value p(x) = 0.

Pearson classified a variety of different shapes and types of distributions according to the nature of the roots of the equation

Although the system is well known and widely available in the literature, there isn’t a clear systematic basis to it. However, all the different types correspond to different forms of solution to the above equations. Pearson‘s first papers [2] covered Types I, III, IV, V, and VI. Later [3], he introduced further special cases and subtypes (VII through XII).

For example, if c1 = c2=0, the differential equation becomes


And K is a constant chosen to make the area under the curve equal to 1, i.e.,

From here, we can deduce that c0 must be positive and

This corresponds to a normal distribution with an expected value (i.e., a mean) of -a and standard deviation √c0. The normal distribution is now viewed as the limit of type I, III, IV, V, or VI.


[1] Johnson, Kotz, and Balakrishnan, (1994), Continuous Univariate Distributions, Volumes I and II, 2nd. Ed., John Wiley and Sons.

[2] Pearson, Karl (1895). “Contributions to the mathematical theory of evolution, II: Skew variation in homogeneous material” (PDF). Philosophical Transactions of the Royal Society. 186: 343–314. Bibcode:1895RSPTA.186..343Pdoi:10.1098/rsta.1895.0010JSTOR 90649.

[3] Pearson, Karl (1901). “Mathematical contributions to the theory of evolution, X: Supplement to a memoir on skew variation”. Philosophical Transactions of the Royal Society A. 197 (287–299): 443–459. Bibcode:1901RSPTA.197..443Pdoi:10.1098/rsta.1901.0023JSTOR 90841.