logo

Mean and Variance of the F-distribution 📂Probability Distribution

Mean and Variance of the F-distribution

Formulas

XF(r1,r2)X \sim F ( r_{1} , r_{2}) Surface Area E(X)=r2r22,r2>2Var(X)=2d22(d1+d22)d1(d22)2(d24),r2>4 E(X) = {{ r_{2} } \over { r_{2} - 2 }} \qquad , r_{2} > 2 \\ \Var(X) = {{ 2 d_{2}^{2} (d_{1} + d_{2} - 2) } \over { d_{1} (d_{2} -2)^{2} (d_{2} - 4) }} \qquad , r_{2} > 4

Derivation

Strategy: Like the chi-squared distribution, the F-distribution also has known moment-generating functions, which we will use.

Moments of the F-Distribution: Let’s say XF(r1,r2)X \sim F(r_{1} , r_{2}) and it can be expressed as X=X1X2\displaystyle X = {{ X_{1} } \over { X_{2} }}. If X1X_{1} and X2X_{2} both follow chi-squared distributions with degrees of freedom d1,d2d_{1}, d_{2} and if d2>2kd_{2} > 2k then the kk-th moment exists EXk=(r2r1)kEX1kEX2k EX^{k} = \left( {{ r_{2} } \over { r_{1} }} \right)^{k} E X_{1}^{k} E X_{2}^{-k}

Moments of the Chi-Squared Distribution: Let’s say Xχ2(r)X \sim \chi^{2} (r). If k>r/2k > - r/ 2 then the kk-th moment exists EXk=2kΓ(r/2+k)Γ(r/2) E X^{k} = {{ 2^{k} \Gamma (r/2 + k) } \over { \Gamma (r/2) }}

Mean

Assuming r2>2r_{2} > 2 then because k=1>r2/2-k = -1 > -r_{2} / 2 thus, EX21EX_{2}^{-1} exists.

If k=1k=1 then according to the moment-generating functions EX1=(r2r1)1EX11EX21=r2r121Γ(r1/2+1)Γ(r1/2)21Γ(r2/21)Γ(r2/2)=r2r12r11r2/21=r2r22 \begin{align*} EX^{1} =& \left( {{ r_{2} } \over { r_{1} }} \right)^{1} E X_{1}^{1} E X_{2}^{-1} \\ =& {{ r_{2} } \over { r_{1} }} {{ 2^{1} \Gamma (r_{1}/2 + 1) } \over { \Gamma (r_{1}/2) }} {{ 2^{-1} \Gamma (r_{2}/2 -1 ) } \over { \Gamma (r_{2}/2) }} \\ =& {{ r_{2} } \over { r_{1} }} 2r_{1} {{ 1 } \over { r_{2}/2 - 1 }} \\ =& {{ r_{2} } \over { r_{2} - 2 }} \end{align*}

Variance

Assuming r2>4r_{2} > 4 then because k=2>r2/2-k = -2 > -r_{2} /2 thus, EX22E X_{2}^{-2} exists.

If k=2k=2 then according to the moment-generating functions EX2=(r2r1)2EX12EX22=(r2r1)222Γ(r1/2+2)Γ(r1/2)22Γ(r2/22)Γ(r2/2)=(r2r1)2(r1/2+1)r1/2(r2/21)(r2/22)=(r2r1)2(r1+2)r1(r22)(r24)=r22(r1+2)r1(r22)(r24) \begin{align*} EX^{2} =& \left( {{ r_{2} } \over { r_{1} }} \right)^{2} E X_{1}^{2} E X_{2}^{-2} \\ =& \left( {{ r_{2} } \over { r_{1} }} \right)^{2} {{ 2^{2} \Gamma (r_{1}/2 + 2) } \over { \Gamma (r_{1}/2) }} {{ 2^{-2} \Gamma (r_{2}/2 -2 ) } \over { \Gamma (r_{2}/2) }} \\ =& \left( {{ r_{2} } \over { r_{1} }} \right)^{2} {{ (r_{1}/2+1)r_{1}/2 } \over { (r_{2}/2-1) (r_{2}/2-2) }} \\ =& \left( {{ r_{2} } \over { r_{1} }} \right)^{2} {{ (r_{1}+2)r_{1} } \over { (r_{2}-2) (r_{2}-4) }} \\ =& {{ r_{2}^{2} (r_{1}+2) } \over { r_{1} (r_{2}-2) (r_{2}-4) }} \end{align*} Therefore Var(X)=r22(r1+2)r1(r22)(r24)[r2r22]2=r22r1(r22)2(r24)[(r1+2)(r22)r1(r24)]=2r22(r1+r22)r1(r22)2(r24) \begin{align*} \Var(X) =& {{ r_{2}^{2} (r_{1}+2) } \over { r_{1} (r_{2}-2) (r_{2}-4) }} - \left[ {{ r_{2} } \over { r_{2} - 2 }} \right]^{2} \\ =& {{ r_{2}^{2} } \over { r_{1} (r_{2} -2)^{2} (r_{2} - 4) }} \left[ (r_{1} + 2)(r_{2} - 2) - r_{1}(r_{2} - 4) \right] \\ =& {{ 2 r_{2}^{2} (r_{1} + r_{2} - 2) } \over { r_{1} (r_{2} -2)^{2} (r_{2} - 4) }} \end{align*}