STA 115 Introduction to Probability Theory 1 ret Trg aes Sampling Distribution 1 Prof. O.1. Osowole Department of Statistics University of Ibadan December 19, 2022 EeAt the end of the lecture, each student should be able to: @ understand sampling distribution with and without replacement. e solve problems associated with sampling distributions.mpling Distribution Statistical inference is the whole process of making a deductive statement about a population from the information obtained from its representative sample. @ Statistical inference therefore involves the process of estimation and test of hypothesis. @ However, most of the conclusions in statistical inference are based on some statistical theory, which are often assumed without much explanation. @ This important concept, which makes it possible to make inference from a sample to its related population, is known as sampling distBefore we give the de! essential to define some basic statistical terms. ition of sampling distribution, it is © Population: is the whole item or things under investigation from which a representative sample is to be taken. Let us denote the population size by N. © Sample:is a representative part of a population. Let us denote the sample size by n . © Statistic:is the quantity calculated from the values of the observations in the sample. e.g. Mean, proportion, median and variance. e@ Parameter population. is a population value or an estimate of theSampling Distribution of Single Mean The probability of the sample statistic is referred to as the sampling distribution of the sample statistic. To illustrate this concept, we shall use one of the most commonly used measures of central tendency, the arithmetic mean (x) The procedure involves taking all possible samples of sizes n from a given population, calculate the mean of each sample and summarize the results into a frequency distribution of the means of the samples. From the sampling distribution we then calculate the mean and variance. We shall then compare our result with the population value. B| Sampling Distribution of Single Mean Suppose we have a population of size N = 6 ie S = {3,5,6,8,2, 9}. The population mean 3+5+6+84+2+9 6 aS ==55 6SSFYanl >) iay2a Daal oleh ace la Meme) ala -M eT a) The population variance is defined as _ @=5.5)? +(5—5.5)? +... + (9 — 5.5)? 6 — 6.25) Now, suppose we take all possible sample size of size n=2 (without replacement). Each element in the population has equal chance of being selected. pteExample 1 contd Selected sample and their means Sample | = 35 | 40 36 | 45 3.8 | 5.5 32 | 25 3,9 | 6.0 5.6 | 55 58 | 65 5.2 | 3.5 5.9 | 7.0 68 | 7.0 62 | 4.0 69 | 75 8.2 | 5.0 8,9 85 © | BD Sampling Distribution of mean for n=2Example 1 contd Selected sample and their means x _| Ff | PCS) | PCR) | PCR) ae || e aa Bie || i] a ag 40|2| # & ie 45|1| 6 | 38 | 4 soli) & ea 2mm 55(/3| & | 4 Ss 60/1| & z 65|1| 6 | SF | “38 70|2| 2 ae ee 75 (1) & e ae asji| 2 | % | 3B 5 | 32.75 ae.Using the idea of mathematical expectation, we know that, = >> #P(x) maa _ 82.5 _ -=- E( 5.5: V(%) = 08 = E (5?) — [E(®)? SOAR) — (SD PR)? =a = 32.75 — 30.25 Ceslution contd Comparing above results with the population value, we observe that:Solution contd It should be noted however that the factor (N=) is called the finite population-correcting factor, which approaches unity as N becomes large. This is especially true when sampling is done without replacement.Suppose we take a sample of size n=4 from the same population in example 1; we have Table: 4. Selected samples and their meansExample 2 contd Sampling distribution of the means =| f | P(x) | P(x) | X2P(2) a0 [il & a 3 45 |1| a ae 475 |2| ¢ 2 so |a| # ae 5.25 |1| + ae 55 [3 = 168 575 [|1| ¢ ae 60 |1| 4 go 6.25 |2| ~ 28 65 [1] & zo |1| # So TotalExample 2 contd Using mathematical expectation again, we could see that E(x) = =55 V(x) = 02 = E(5?) — [E(2)? = SD P(%) — (SD PRP = 30.875 — 30.25 o2 = 0.625 mitComparing above results with the population value (n=4) Therefore, S.E(X) = V/0.625 = 0.7905.@)_ We could observe from examples 1 and 2, that the mean of the sampling distribution and that of the population when all possibles samples n=2 and n: were taken are the same while the variance reduced as sample size increases. This canfirms the law of large numbers that says "as sample size increases the variance or standard error of an estimate tends to zero” ,that is S.E(R) 0 as n> (i) Since lin oo) > 2 We observe that as the sample size increases, Le (7 then 30),oka