A Very Brief Summary of Statistical Inference, and Examples

Transcription

1 A Very Brief Summary of Statistical Inference, and Examples Trinity Term 2009 Prof. Gesine Reinert Our standard situation is that we have data x = x 1, x 2,..., x n, which we view as realisations of random variables X 1, X 2,..., X n with a distribution (model) f(x 1, x 2,..., x n ; θ), where θ is unknown. In frequentist analysis, θ Θ is an unknown constant. as 1. Likelihood and Sufficiency (Fisherian) Likelihood approach: We define the likelihood of θ given the data L(θ) = L(θ, x) = f(x 1, x 2,..., x n ; θ). Often: X 1, X 2,..., X n are independent, identically distributed (i.i.d.); then L(θ, x) = n i=1 f(x i, θ). We summarize information about θ: find a minimal sufficient statistic t(x); from the Factorization Theorem: T = t(x) is sufficient for θ if and only if there exists functions g(t, θ) and h(x) such that for all x and θ f(x, θ) = g(t(x), θ)h(x). Moveover T = t(x) is minimal sufficient when it holds that f(x, θ) f(y, θ) is constant in θ t(x) = t(y) Example Let X 1,..., X n be a random sample from the truncated exponential distribution, where f Xi (x i ) = e θ xi, x i > θ or, using the indicator function notation, f Xi (x i ) = e θ xi I (θ, ) (x i ). Show that Y 1 = min(x i ) is sufficient for θ. Let T = T (X 1,..., X n ) = Y 1. We need the pdf f T (t) of the smallest order statistic. Calculate the cdf for X i, Now F (x) = x θ e θ z dz = e θ [e θ e x ] = 1 e θ x. n P (T > t) = (1 F (t)) = (1 F (t)) n. i=1 1

2 Differentiating gives that f T (t) equals n[1 F (t)] n 1 f(t) = ne (θ t)(n 1) e θ t = ne n(θ t), t > θ. So the conditional density of X 1,... X n given T = t is e θ x1 e θ x2... e θ xn ne n(θ t) = e xi ne nt, x i t, i = 1,... n, which does not depend on θ for each fixed t = min(x i ). Note that since x i t, i = 1,..., n, neither the expression nor the range space depends on θ, so the first order statistic, X (1) = min(x i ), is a sufficient statistic for θ. With Alternatively, use the Factorization Theorem: f X (x) = n e θ xi I (θ, ) (x i ) i=1 = I (θ, ) (x (1) ) n i=1 e θ xi = e nθ I (θ, ) (x (1) )e n i=1 xi. g(t, θ) = e nθ I (θ, ) (t) and h(x) = e n i=1 xi we see that X (1) is minimal sufficient. 2. Point Estimation Estimate θ by a function t(x 1,..., x n ) of the data; often by maximumlikelihood: ˆθ = arg max L(θ), θ or by method of moments. Neither are unbiased in general, but the m.l.e. is asymptotically unbiased and asymptotically efficient; under some regularity assumptions, ˆθ N (θ, In 1 (θ)), where [ ( l(θ, ) ] 2 X) I n (θ) = E θ is the Fisher information (matrix). Under more regularity, ( 2 ) l(θ, X) I n (θ) = E θ 2. If x is a random sample, then I n (θ) = ni 1 (θ) ; often we abbreviate I 1 (θ) = I(θ). 2

3

4 Note that if c was large, then by the CLT each observation X j is approximately normal, because, in distribution, X j = Y 1 + Y Y k, with Y i, i = 1,..., k, being i.i.d. Gamma(c/k, β). Recall also that there is an iterative method to compute m.l.e.s, related to the Newton-Raphson method. 3. Hypothesis Testing For simple null hypothesis and simple alternative, the Neyman-Pearson Lemma says that the most powerful tests are likelihood-ratio tests. These can sometimes be generalized to one-sided alternatives in such a way as to yield uniformly most powerful tests. Example Suppose as above that X 1,..., X n is a random sample from the truncated exponential distribution, where f Xi (x i ; θ) = e θ xi, x i > θ. Find a UMP test of size α for testing H 0 : θ = θ 0 against H 1 : θ > θ 0 : Let θ 1 > θ 0, then the LR is f(x, θ 1 ) f(x, θ 0 ) = e n(θ1 θ0) I (θ1, )(x (1) ), which increases with x (1), so we reject H 0 if the smallest observation, T = X (1), is large. Under H 0, we have calculated that the pdf of f T is So for t > θ, under H 0, P (T > t) = ne n(θ0 y1), y 1 > θ 0. For a test at level α, choose t such that t ne n(θ0 s) ds = e n(θ0 t). t = θ 0 ln α n. The LR test rejects H 0 if X (1) > θ 0 ln α n. The test is the same for all θ > θ 0, so it is UMP. For a general null hypothesis and a general alternative, H 0 : θ Θ 0 versus H 1 : θ Θ 1 = Θ\Θ 0, the (generalized) LR test uses the likelihood ratio statistic T = max L(θ; X) θ Θ max L(θ; X). θ Θ 0 4

5 We reject H 0 for large values of T ; we use the chisquare asymptotics 2 log T χ 2 p, where p = dimθ dimθ 0 (which is valid for nested models only). Alternatively, we could use score tests, which are based on the score function l/ θ, with asymptotics in terms of normal distribution N (0, I n (θ)). Or, we could use a Wald test, which is based on the asymptotic normality of the m.l.e.; often ˆθ N ( θ, In 1 (θ) ) if θ is the true parameter. Recall: In a random sample, I n (θ) = ni 1 (θ). An important example is Pearson s Chisquare test, which we derived as a score test, and we saw that it is asymptotically equivalent to the generalized likelihood ratio test Example Let X 1,..., X n be a random sample from a geometric distribution with parameter p; P (X i = k) = p(1 p) k, k = 0, 1, 2,.... Then and so that and L(p) = p n (1 p) (x i) l(p) = n ln p + nx ln(1 p), ( 1 l/ p(p) = n p x ) 1 p ( 2 l/ p 2 (p) = n 1 ) p 2 x (1 p) 2. We know that EX 1 = (1 p)/p. Calculate the information I(p) = n p 2 (1 p). Suppose n = 20, x = 3, H 0 : p 0 = 0.15 and H 1 : p Then l/ p(0.15) = 62.7 and I(0.15) = The test statistic is then Z = 62.7/ = Compare to 1.96 for a test at level α = 0.05: do not reject H 0. Example continued. For a generalized LRT the test statistic is based on 2 log[l(ˆθ)/l(θ 0 )] χ 2 1, and the test statistic is thus 2(l(ˆθ) l(θ 0 ) = 2n(ln(ˆθ) ln(θ 0 ) + x(ln(1 ˆθ) ln(1 θ 0 )). We calculate that the m.l.e. is ˆθ = x. 5

6 Suppose again n = 20, x = 3, H 0 : θ 0 = 0.15 and H 1 : θ The m.l.e. is ˆθ = 0.25, and l(0.25) = 44.98; l(0.15) = Calculate χ 2 = 2( ) = 5.4 and compare to chisquare distribution with 1 degree of freedom: 3.84 at 5 percent level, so reject H Confidence Regions If we can find a pivot, a function t(x, θ) of a sufficient statistics whose distribution does not depend on θ, then we can find confidence regions in a straightforward manner. Otherwise we may have to resort to approximate confidence regions, for example using the approximate normality of the m.l.e. Recall that confidence intervals are equivalent to hypothesis tests with simple null hypothesis and one- or two-sided alternatives Not to forget about: Profile likelihood Often θ = (ψ, λ) where ψ contains the parameters of interest; then we may base our inference on the profile likelihood for ψ, L P (ψ) = L(ψ, ˆλ ψ ). Again we can use a (generalized) LRT or score test; if ψ scalar, H 0 : ψ = ψ 0, H + 1 : ψ > ψ 0, H 1 : ψ < ψ 0, use test statistic T = l(ψ 0, ˆλ 0 ; X), ψ where ˆλ 0 is the MLE for λ when H 0 true. Large positive values of T indicate H + 1 ; large negative values indicate H 1, and T l ψ I 1 λ,λ I ψ,λl λ N(0, 1/I ψ,ψ ), where I ψ,ψ = (I ψ,ψ I 2 ψ,λ I 1 λ,λ ) 1 is the top left element of I 1. We estimate the parameters by substituting the null hypothesis values; calculate the practical standardized form of T as Z = T Var(T ) l ψ(ψ, ˆλ ψ )[I ψ,ψ (ψ, ˆλ ψ )] 1/2, which is approximately standard normal. 6

7 Bias and variance approximations: the delta method The delta method is useful if we cannot calculate mean and variance directly. Suppose T = g(s) where ES = β and Var S = V. Taylor expansion gives T = g(s) g(β) + (S β)g (β). Taking the mean and variance of the r.h.s.: ET g(β), Var T [g (β)] 2 V. This also works for vectors S, β, with T still a scalar. If ( g (β) ) i = g/ β i, and g (β) the matrix of second derivatives, then Var T [g (β)] T V g (β) and ET g(β) trace[g (β)v ]. Exponential family For distributions in the exponential family, many calculations have been standardized, see the lecture notes. 7