Definition 1: The chi-square distribution with k degrees of freedom, abbreviated χ2(k), has probability density function
k does not have to be an integer and can be any positive real number.
Click here for more technical details about the chi-square distribution, including proofs of some of the propositions described below. Except for the proof of Corollary 2 knowledge of calculus will be required.
Observation: The chi-square distribution is the gamma distribution where α = k/2 and β = 2.
Property 1: The χ2(k) distribution has mean k and variance 2k
Observation: The key statistical properties of the chi-square distribution are:
- Mean = k
- Median = k-2⁄3 for large k
- Mode = k – 1 for k > 2
- Range = [0.∞)
- Variance = 2k
- Skewness =
- Kurtosis = 12/k
The following are the graphs of the pdf with degrees of freedom df = 5 and 10. As df grows larger the fat part of the curve shifts to the right and becomes more like the graph of a normal distribution.
Figure 1 – Chart of chi-square distributions
Theorem 1: Suppose x has standard normal distribution N(0, 1) and let x1, …, xk be k independent sample values of x, then the random variable
has a chi-square distribution χ2(k).
- If x has distribution N(0, 1) then x2 has distribution χ2(1)
- If x ~ N(μ, σ) and z = (x–μ)/σ then over repeated samples z2 has distribution χ2(1)
- If x1, …, xk are independent observations from a normal population with distribution N(μ,σ) and for each i, z = (x–μ)/σ , then the following random variable has a χ2(k) distribution
Property 2: If x and y are independent and x has distribution χ2(m) and y has distribution χ2(n), then x + y has distribution χ2(m + n)
Theorem 2: If x is drawn from a normally distributed population N(μ, σ) then for samples of size n the sample variance s2 has distribution
Corollary 2: s2 is an unbiased, consistent estimator of the population variance
Corollary 3: If x is drawn from a normally distributed population N(μ, σ), then for samples of size n the random variable has a χ2(n–1), distribution
Property 3: The mean of the sample variance s2 is σ2 and the variance is
Proof: This can be seen from the proof of Corollary 2.
Excel Functions: Excel provides the following functions:
CHIDIST(x, df) = the probability that the chi-square distribution with df degrees of freedom is ≥ x; i.e. 1 – F(x) where F is the cumulative chi-square distribution function.
CHIINV(α, df) = the value x such that CHIDIST(x, df) = 1 – α; i.e. the value x such that the right tail of the chi-square distribution with area α occurs at x. This means that F(x) = 1 – α, where F is the cumulative chi-square distribution function.
With Excel 2010/2013 there are a number of new functions (CHISQ.DIST, CHISQ.INV, CHISQ.DIST.RT and CHISQ.INV.RT) that provide equivalent functionality to CHIDIST and CHIINV, but whose syntax is more consistent with other distribution functions. These functions are described in Built-in Statistical Functions.
In Excel 2010 CHISQ.DIST(x, df, TRUE) is the cumulative distribution function for the chi-square distribution with df degrees of freedom, i.e. 1 – CHIDIST(x, df), and CHISQ.DIST(x, df, FALSE) is the pdf for the chi-square distribution.
Real Statistics Functions: The Real Statistics Resource Pack provides the following supplemental functions.
CHISQ_DIST(x, df, cum) = GAMMA.DIST(x, df/2, 2, cum) = GAMMADIST(x, df/2, 2, cum)
CHISQ_INV(p, df) = GAMMA.INV(p, df/2, 2) = GAMMAINV(p, df/2, 2)
These functions provide better estimates of the chi-square distribution when df is not an integer. The first function is also useful in providing an estimate of the pdf for versions of Excel prior to Excel 2010, where CHISQ.DIST(x, df, FALSE) is not available.
Example 1: Suppose we take samples of size 10 from a population with normal distribution N(0, 2). Find the mean and variance of the sample distribution of s2.