The X2 Distribution.doc

The Χ2 Distribution In the same manner as the normal (σ known) and t-distribution (σ unknown) are the sampling distributions for the mean, the χ2 (pronounced Chi-squared) is the sampling distribution for the variance/standard deviation. Since the variance/standard deviation must be positive, it is clear that both the normal and t-distributions are redundant (since both Z and t-distributions can take on negative values) However, if we square Z we obtain an entirely positive horizontal axis (0 to + ∞). This is the χ2 distribution with 1 (one) degree of freedom. Z2 = χ2(1) Like the t-distribution, a sample mean has had to be calculated, as such, we lose a degree of freedom, thus our degrees of freedom (υ) are the number of observations (n) – 1. The shape (specifically the skew), of the χ2 distribution depends on the degrees of freedom. At low DoF the distribution leans to the left (annoyingly, this is called a right skew) but becomes more symmetric and approaches the normal distribution as DoF increase. χ2 Like all probability distributions, the area under the curve must equal 1 (one) The mean of the distribution is υ, the number of DoF. The variance is given as 2υ. The χ2 statistic is given as follows: Where: S2 = Sample Variance σ2 = Population Variance υ = n-1 Note that the sum of two independent distributed random variables χ2(υ=α) and χ2(υ=β) is itself distributed χ2(υ = α + β) Like the t-distribution, χ2 tables depend on the DoF. The first column of the tables, “υ”, lists the DOF, the first row denotes the area to the right of a critical, or critical, values. For example. For υ = 10, 90% of the observation of a χ2 distributed random variable fall to the left of χ210 = 18.307 obviously then, 10% of observation fall to the right For υ = 20, 95% of the observations of a χ2 distributed random variable fall between χ220 = 9.591 and χ220 = 34.170 For υ = 40, 99% of the observations of a χ2 distributed random vari