INTERVAL ESTIMATION and HYPOTHESES TESTING 1.1. Probability Distributions. Table 1 Contains Information on Some Common Probabili

INTERVAL ESTIMATION and HYPOTHESES TESTING 1.1. Probability Distributions. Table 1 Contains Information on Some Common Probabili

INTERVAL ESTIMATION AND HYPOTHESES TESTING 1. BRIEF REVIEW OF SOME COMMON DISTRIBUTIONS 1.1. Probability Distributions. Table 1 contains information on some common probability distri- butions. TABLE 1. Probability Distributions Distribution pdf Mean Variance mgf Notes 2 2 2 µ =0,σ =1 −1 x−µ 2 t σ −1 2 µt+ 2 1 x Normal, N(µ, σ2) √ 1 · e 2 σ µ σ2 e 2 f(x;µ,σ )= √ ·e 2 2πσ2 2π t2 MX (t)=e 2 1 −x − · θ 2 − 1 Exponential θ e ,θ>0 θ θ (1 θt) Xi ∼N(0, 1),i=1, 2, ..., n n 2 2 ⇒ X Xi ∼ χ (n) i=1 2 − − − Xi ∼N (µ, σ ),i=1, 2, ..., n 1 ν 2 x ν χ2(ν) x 2 e 2 ν 2ν (1 − 2t) 2 Chi-square, ν Γ v n 2 2 2 ( 2 ) X − µ ⇒ X i ∼ χ2(n) σ i=1 n X − X¯ 2 ⇒ X i ∼ χ2(n − 1) σ i=1 −x 1 α−1 β 2 − −α Gamma βαΓ(α) x e αβ αβ (1 βt) N(0,1) ν +1 −(ν +1) t(ν)= Γ 2 2 r 2 √ 2 x ν χ (ν) t(ν) ν 1+ ν 0 ν − 2 ν πν Γ( 2 ) N(0,1) and χ2(ν) are independent ν +ν Γ 1 2 ν1 ν χ2(ν ) 2 ν 1 −1 1 1 · 1 2 · 2 χ2 (ν ) ν1 ν2 ν x 2 ν1 ν2 1 1 Γ Γ 2 ν 2ν (ν +ν −2) F (ν1 ,ν2 )= = · ( 2 ) ( 2 ) 2 2 1 2 χ2(ν ) ν1 χ2 (ν ) F(ν1 ,ν2) 2 2 2 2 2 ν2 −2 ν1 (ν2−2) (ν2−4) −(ν1 +ν2) ν2 ν1 2 2 ×1+ x 2 χ (ν1 ) and χ (ν2 ) are independent ν2 1 1 1.2. Moments. 1.2.1. Population raw moments. ∞ 0 r Z r µr = E(X )= x f(x) dx −∞ ∞ (1) 0 Z µ1 = E(X)= xf(x) dx −∞ Date: December 8, 2005. 1 2 INTERVAL ESTIMATION AND HYPOTHESES TESTING 1.2.2. Population central moments. ∞ r Z r µr = E[(X − µX ) ]= (x − µX ) f(x) dx −∞ ∞ Z µ1 = E[X − µX ]= (x − µX ) f(x) dx (2) −∞ ∞ 2 Z 2 2 µ2 = E[(X − µX ) ]= (x − µX ) f(x) dx = Var(X)=σ −∞ 1.2.3. Sample raw moments. 1 n X¯ r = X Xr n n i i=1 (3) 1 n X¯ = X X n i i=1 1.2.4. Sample central moments. n r 1 0 r C = X Xi − µ r =1, 2, 3,..., n n i,1 i=1 n 1 1 0 C = X Xi − µ (4) n n i,1 i=1 n 2 1 0 2 C = X Xi − µ n n i,1 i=1 1.2.5. Sample moments about the average (sample mean). n r 1 ¯ r M = X Xi − Xn ,r=1, 2, 3,..., n n i=1 n 1 ¯ Mn = X Xi − Xn =0 (5) n i=1 n 2 1 ¯ 2 2 M = X Xi − Xn =ˆσ n n i=1 1.3. Properties of various sample moments and functions thereof. Consider a sample X1,X2, ... 2 ,Xn where the Xi are identically and independently distributed with mean µ and variance σ . Now consider a sample X1,X2, ... ,Xn where the Xi are identically and independently dis- tributed normal random variables with mean µ and variance σ2. 2. THE BASIC IDEA OF INTERVAL ESTIMATES An interval rather than a point estimate is often of interest. Confidence intervals are thus impor- tant in empirical work. To construct interval estimates, standardized normal random variables are often used. INTERVAL ESTIMATION AND HYPOTHESES TESTING 3 TABLE 2. Properties of sample moments Moment Expected Value Variance ¯ 1 n σ2 X = n Pi=1 Xi µ n 1 1 n − σ2 Cn = n Pi=1 (Xi µ) 0 n 2 E(X−µ)4 −[E(X−µ)2 ] 4 2 1 n − 2 2 1 − 2 µ4−σ Cn = n Pi=1 (Xi µ) σ n Var(X µ) = n = n n 2 − 2 − − 4 µ −µ2 2(µ −2µ2) µ −3µ2 2 1 − ¯ n−1 2 (n 1) µ4 − (n 1)(n 3)σ 4 2 − 4 2 4 2 Mn = n Pi=1 Xi X n σ n3 n3 = n n2 + n3 2 4 2 1 n − ¯ 2 µ4 − (n−3)σ S = n−1 Pi=1 Xi X σ n n(n − 1) TABLE 3. Properties of sample moments of a normal distribution Moment Value Expected Value Variance µ1 = E(X − µ) 0 - - 2 2 µ2 = E(X − µ) σ - - 3 µ3 = E(X − µ) 0 - - 4 4 µ4 = E(X − µ) 3σ - - ¯ 1 n σ2 X = n Pi=1 Xi - µ n 2 1 n − ¯ 2 2 2σ4 S = n−1 Pi=1 Xi X - σ n−1 4 2 2 1 n − ¯ 2 n−1 2 2σ (n−1) Mn =ˆσ = n Pi=1 Xi X - n σ n2 3. STANDARDIZED NORMAL VARIABLES AND CONFIDENCE INTERVALS FOR THE MEAN WITH σ KNOWN 3.1. Form of the confidence interval. If X is a normal random variable with mean µ and variance σ2 then X − µ Z = (6) σ is a standard normal variable with mean zero and variance one. An estimate of µ, µˆ is given by the sample mean X¯. This then implies 1 n σ2 X¯ = X X = X¯ ∼ N µ, n i n i=1 (7) X¯ − µ z = ∼ N(0, 1) h √σ i n So if γ1 is the upper α/2 percent critical value of a standard normal variable, i.e. 4 INTERVAL ESTIMATION AND HYPOTHESES TESTING ∞ Z 1 − 1 z2 α √ e 2 dz = then γ1 2π 2 " x¯ − µ # 1 − α = F (γ1) − F (−γ1)=Pr −γ1 ≤ ≤ γ1 √σ n σ σ =Pr−γ √ ≤ x¯ − µ ≤ γ √ 1 n 1 n (8) σ σ =Prγ √ ≥−x¯ + µ ≥−γ √ 1 n 1 n σ σ =Prx¯ − γ √ ≤ µ ≤ x¯ + γ √ 1 n 1 n Therefore, with σ known, σ σ x¯ − γ √ ≤ µ ≤ x¯ + γ √ (9) 1 n 1 n is said to be the (1 − α) 100% confidence interval for µ. 3.2. Example confidence interval for the mean of a normal distribution with σ2 known. Consider the income data for carpenters and house painters in table 4 where the data for both groups of individuals is distributed normally. Further assume that σ2 is known and equal to $600,000. TABLE 4. Income Data for Carpenters and House Painters carpenters painters sample size nc =12 np =15 mean income c¯ = $6000 p¯ = $5400 2 2 estimated variance sc = $565 000 sp = $362 500 Consider a 95% confidence interval for the mean of the distribution which we will denote by µc. α For the normal distribution with 2 = 0.025, γ1 is equal to 1.96. We then have √ √ " 600 000 600 000 # 1 − α =Pr 6000 − √ (1.96) ≤ µc ≤ 6000 + √ (1.96) 12 12 (10) = Pr [6000 − (223.607)(1.96) ≤ µx ≤ 6000 + (223.607)(1.96)] = Pr [5561.73 ≤ µx ≤ 6438.27] 4. CONFIDENCE INTERVALS FOR THE MEAN WITH σ UNKNOWN 4.1. Form of the confidence interval. The previous section gave an interval estimate for the mean of a population when σ was known. When σ is unknown, another method must be used. Recall from the section 1 on probability distributions that the t random variable is defined as INTERVAL ESTIMATION AND HYPOTHESES TESTING 5 z t = (11) q χ2(ν) ν where z is a standard normal and χ2(ν) is a χ2 random variable with ν degrees of freedom and z and χ2(ν) are independent. 2 Also note from the same section 1 that if Xi ∼ N(µ, σ ) then n X − µ 2 X i ∼ χ2(n) σ i=1 (12) n ¯ 2 Xi − X 2 and X ∼ χ (n − 1) σ i=1 If Y ∼ N(β, σ2) and Y¯ = βˆ, then we would have n ˆ 2 Yi − β ! X ∼ χ2(n − 1) (13) σ i=1 (n−1)S2 2 We can use the information in equation 12 to find the distribution of σ2 where S is equal to 1 n S2 = X(X − X¯)2 (14) n − 1 i i=1 2 2 (n−1)S 2 Now substitute for S in σ2 where S and simplify as follows. (n − 1)S2 (n − 1) 1 n = · X(X − X¯)2 σ2 σ2 n − 1 i i=1 1 n = · X(X − X¯)2 σ2 i (15) i=1 n X − X¯ 2 = X i ∼ χ2(n − 1) σ i=1 The last line then indicates that (n − 1)S2 ∼ χ2(n − 1) (16) σ2 X¯ −µ (n−1)S2 We can show that in equation 7 and 2 in equation 16 are independent so that h √σ i σ n 6 INTERVAL ESTIMATION AND HYPOTHESES TESTING X¯ − µ X¯ − µ σ σ √ √ n n = 1 √ s(n − 1)S2 S2 σ σ2(n − 1) (17) √ X¯ − µ n X¯ − µ = = ∼ t(n − 1) S √ S n If γ1 is the upper α/2 percent critical value of a t random variable then ν +1 (ν+1) ∞ Γ 2 − Z 2 t 2 α 1+ dt = (18) √ ν ν 2 γ1 πν Γ 2 and X¯ − µ σ √ n 1 − α = F (γ1) − F (−γ1)=Pr−γ1 ≤ ≤ γ1 2 s(n − 1)S σ2(n − 1) √ n(X¯ − µ) (19) =Pr−γ ≤ ≤ γ 1 S 1 −γ S γ S =Pr √1 ≤ X¯ − µ ≤ √1 n n γ S γ S =PrX¯ − √1 ≤ µ ≤ X¯ + √1 n n This is referred to as (1−α)(100%) confidence interval for µ.

View Full Text

Details

  • File Type
    pdf
  • Upload Time
    -
  • Content Languages
    English
  • Upload User
    Anonymous/Not logged-in
  • File Pages
    29 Page
  • File Size
    -

Download

Channel Download Status
Express Download Enable

Copyright

We respect the copyrights and intellectual property rights of all users. All uploaded documents are either original works of the uploader or authorized works of the rightful owners.

  • Not to be reproduced or distributed without explicit permission.
  • Not used for commercial purposes outside of approved use cases.
  • Not used to infringe on the rights of the original creators.
  • If you believe any content infringes your copyright, please contact us immediately.

Support

For help with questions, suggestions, or problems, please contact us