Fundamental Sampling DistributionsLinear Combinations of R.V.’sSlide 3Slide 4SamplingSlide 6EstimatorsSampling DistributionsSampling Distributions - SetupSampling Distributions - MeanSlide 11Slide 12Central Limit Theorem (CLT)Sampling Distributions - MeansSlide 15Slide 16Slide 17Sampling Distributions - VarianceSlide 19Slide 20Slide 21Student’s t-DistributionSlide 23Slide 24F-DistributionSlide 26Slide 27Slide 28Upper Tail Probabilities in ExcelUpper Tail Percentiles in Excel1Fundamental Sampling DistributionsLinear Combinations of R.V.’sSamplingEstimators•Central tendency•Variability or dispersionSampling Distributions•Sample mean and sample variance•Student’s t-distribution•F-distribution2Linear Combinations of R.V.’sX1, X2, … , Xn are independent and normally distributed r.v.’s with means 1, 2, … , n and standard deviations 1, 2, … , n , respectivelyLet Y = a1X1 + a2X2 + … + anXn•Y is normally distributed• • nnaaaYE 2211][2222222121)(nnaaaYV 3Linear Combinations of R.V.’sX1, X2, … , Xn are mutually independent r.v.’s distributed chi-squared with v1, v2, … , vn degrees of freedom , respectivelyLet Y = X1 + X2 + … + Xn•Y is distributed chi-squared with degrees of freedom Denoted as 2(v)nvvvv 214Linear Combinations of R.V.’sX1, X2, … , Xn are identically distributed normal r.v.’s with mean and variance 2Let •Y is distributed chi-squared with n degrees of freedom, denoted as 2(n)niiniiZXY12125SamplingParameter•A property of an probability distribution.•In real life, these are typically unknown.Population•The totality of the observations of interest.•Typically, it is not practical to see all observations in a population.Sample•A subset of observations from a population.•These should be representative of the population.6SamplingRandom Sample•Before data: X1, X2, … , Xn are independent and identically distributed (i.i.d.) r.v.’s•After data: Denote observed values by x1, x2, …, xn7EstimatorsA statistic is a function u(X1, X2, … , Xn)An estimator is a statistic designed to estimate an unknown parameter• is an estimator for the population mean • is an estimator for the population median•S2 is an estimator for the population variance 2Statistics and estimators are r.v.’sAn estimate is a calculated value (not random) of an estimator•e.g., , , s2XX~~xx~8Sampling DistributionsBefore we collect data, the value of a statistic u(X1, X2, … , Xn) is randomIf we collect multiple random samples: •Sample 1 •Sample 2 •etc.This variability is called sampling errorA sampling distribution is the distribution for a statistic2.3x9.2x9Sampling Distributions - SetupLet X denote an observation to be sampled•X follows the population distributionIf we collect a sample of n observations•X1, X2, … , Xn are i.i.d. population distribution•E[Xi] = = the population mean•V(Xi) = 2 = the population varianceExample: Suppose we are studying the lifetime of lightbulbs, and we collect n = 100•Assume the population distribution is Exp()•X1, X2, … , X100 are i.i.d. Exp()10Sampling Distributions - MeanSample Mean (before data)Expected value of )(1111111112121nnnXEnXEnXEnXEnXXXnEXEniniinn nniiXXXnXnX 21111X11Sampling Distributions - MeanVariance of•Note: X1, X2, … , Xn are independent r.v.’s nnnnXVnXVnXVnXVnXXXnVXVniniinn222122122221221)(1111111X12Sampling Distributions - MeanIf the population distribution is N( , 2)•X1, X2, … , Xn are i.i.d. N( , 2)• is distributed N( , 2/n)• is distributed N(0 , 1)Example: X1, …, X25 i.i.d. N( = 15 , 2 = 100)XnXZ/ 9938.0)5.2(5/101520/20 nXPXP13Central Limit Theorem (CLT)X1, …, Xn i.i.d some population distribution with mean and variance 2For large n• is approximately distributed N( , 2/n)Example: X1, …, X64 i.i.d. Exp( = 4) = E[Xi] = = 4 , 2 = V(Xi) = 2 = 16X 0228.0)2(128/443/3ZPnXPXP14Sampling Distributions - MeansTwo Independent Populations (any distributions)•Pop. 1 Sample 1 with n1 obsns •Pop. 2 Sample 2 with n2 obsns Expected value of Variance of •Note: and are independent r.v.’s1X2X 212121 XEXEXXE21XX 21XX 2221212121nnXVXVXXV1X2X),(211),(22215Sampling Distributions - MeansIf Pop. 1 is and Pop. 2 is • is distributedCLT: For large n1 and n2 and for any distns, if Pop. 1 has mean 1 and variance , and Pop. 2 has mean 2 and variance • is approximately distributed),(211N),(222N21XX 22212121,nnN212221XX 22212121,nnN16Sampling Distributions - MeansExample: Pop. 1 is distributed Exp( = 4) and Pop. 2 is distributed Gamma( = 3, = 4) 1 = = 4 , 2 = = 12 , Sample 64 obsns from Pop.1 and 75 obsns from Pop.2 1 2 = 8• 89.075486416222121nn162214822217Sampling Distributions - Means 9830.0)12.2(12.289.0)8(6)(6222121212121ZPnnXXPXXP18Sampling Distributions - VarianceSample Variance (before data)•Note:Assume X1, X2, … , Xn are i.i.d. N( , 2)•Then is distributed 2(n 1) 22)1(Sn niiXXnS122)(11 212)1( SnXXnii19Sampling Distributions - VarianceRecall is distributed 2(n)Now
View Full Document