F Distribution and Moment Boundaries

The F-distribution arises as the ratio of two independent chi-square random variables, each divided by its respective degrees of freedom. It plays a fundamental role in analysis of variance (ANOVA) and testing hypotheses about variances.

Worked Example

Let's calculate the F-distribution properties when the numerator degrees of freedom (a) is 4, the denominator degrees of freedom (b) is 5, and we want to evaluate the probability density at x = 1.

Identify parameters: a = 4, b = 5, x = 1.
Calculate the mean: The mean of an F-distribution is defined only when b > 2. It is given by b / (b - 2). Here, b = 5 > 2, so Mean = 5 / (5 - 2) = 5 / 3 ≈ 1.6666666666666667.
Calculate the variance: The variance is defined only when b > 4. It is given by [2b²(a + b - 2)] / [a(b - 2)²(b - 4)]. Here, b = 5 > 4, so Variance = [2(25)(4 + 5 - 2)] / [4(5 - 2)²(5 - 4)] = [50(7)] / [4(9)(1)] = 350 / 36 = 9.722222222222221.
Calculate the density at x = 1. The density f(x) uses the Beta function B(a/2, b/2): f(x) = ( (a/b)^(a/2) * x^(a/2 - 1) ) / ( B(a/2, b/2) * (1 + (a/b)x)^((a+b)/2) ). With a=4, b=5, a/2=2, b/2=2.5. The density at x=1 evaluates to approximately 0.39761407921806946.

Experiment

Observe how the shape and properties of the F-distribution change as you adjust the degrees of freedom.

Open the F Distribution and Moment Boundaries module.
Select Numerator df to 4 and Denominator df to 2. Notice that the mean and variance become undefined because the heavy tail drops off too slowly.
Change the Numerator df to 1 and the Probe x to 0. Notice that the density near x=0 becomes unbounded (it shoots up to infinity).

Limitations

The most important limitation of the F-distribution involves the non-existence of its moments. For small denominator degrees of freedom (b), the distribution's tail is very heavy. Specifically, the mean does not exist if b ≤ 2, and the variance does not exist if b ≤ 4. In these regions, classical statistics that rely on finite means or variances break down. Furthermore, when the numerator degrees of freedom (a) is 1, the density function has a singularity (is unbounded) at x = 0.

References

For more details on the density function, beta representations, and exact moment conditions, refer to the mathematical specifications provided by:

SciPy Statistical Functions: F-distribution (Supports the PDF definitions and moments.)

F Distribution and Moment Boundaries