Sampling Distributions - BMS 511 Biostats & Statistical Analysis PDF

BMS 511 Biostats & Statistical Analysis Chapter 13 Sampling distributions Guang Xu, PhD, MPH Assistant Professor of Biostatistics and Public Health College of Osteopathic Medicine Marian Universi...

BMS 511 Biostats & Statistical Analysis Chapter 13 Sampling distributions Guang Xu, PhD, MPH Assistant Professor of Biostatistics and Public Health College of Osteopathic Medicine Marian University Copyright © 2018 W. H. Freeman and Company  Previous Learning Objectives Demonstrate Normal distributions Normal distributions The 68–95–99.7 rule The standard Normal distribution Using the standard Normal table (Table B) Inverse Normal calculations Normal quantile plots Copyright © 2018 W. H. Freeman and Company  Learning Objectives Demonstrate sampling distributions Parameter versus statistic Sampling distributions Sampling distribution of the sample mean The central limit theorem Sampling distribution of the sample proportion The law of large numbers Copyright © 2018 W. H. Freeman and Company  Parameter versus Statistic Population: the entire group of individuals in which we are interested but usually can’t assess directly. A parameter is a number summarizing the population. Parameters are usually unknown. Sample: the part of the population we actually examine and for which we do have data. A statistic is a number summarizing a sample. We often use a statistic to estimate an unknown population parameter. Copyright © 2018 W. H. Freeman and Company  Sampling distributions Different random samples taken from the same population will give different statistics. But there is a predictable pattern in the long run. A statistic computed from a random sample is a random variable. The sampling distribution of a statistic is the probability distribution of that statistic for samples of a given size n taken from a given population. Copyright © 2018 W. H. Freeman and Company  Sampling distribution of the sample mean The mean of the sampling distribution of is μ. – There is no tendency for a sample average to fall systematically above or below μ, even if the population distribution is skewed.  is an unbiased estimate of the population mean μ. The standard deviation of the sampling distribution of is σ/√n. – The standard deviation of the sampling distribution measures how much the sample statistic varies from sample to sample. Averages are less variable than individual observations. Copyright © 2018 W. H. Freeman and Company  For Normally distributed populations (1 of 2) When a variable in a population is Normally distributed, the sampling distribution of the sample mean x̅ is also Normally distributed. Copyright © 2018 W. H. Freeman and Company  For Normally distributed populations (2 of 2) population N(µ, σ) sampling distribution N(µ, σ/√n) Copyright © 2018 W. H. Freeman and Company  Sampling distribution example (1 of 2) The blood cholesterols of 14-year-old boys is ~ N(µ = 170, σ = 30) mg/dL. The population: The middle 99.7% of cholesterol levels in boys is 80 to 260 mg/dL. Copyright © 2018 W. H. Freeman and Company  Sampling distribution example (2 of 2) Now consider random samples of 25 boys. The sampling distribution of average cholesterol levels is ~ N(µ = 170, σ = 30/√25 = 6) mg/dL: The middle 99.7% of average cholesterol levels (of 25 boys) is 152 to 188 mg/dL. Copyright © 2018 W. H. Freeman and Company  Another sampling distribution example Deer mice (Peromyscus maniculatus) have a body length (excluding the tail) known to vary Normally, with a mean body length µ = 86 mm, and standard deviation σ = 8 mm. https://PollEv.com/guangxu242 Or text guangxu242 to 37607 For random samples of 20 deer mice, the distribution of the sample mean body length is A. Normal, mean 86, standard deviation 8 mm. B. Normal, mean 86, standard deviation 20 mm. C. Normal, mean 86, standard deviation 1.789 mm. D. Normal, mean 86, standard daeviation 3.9 mm. Copyright © 2018 W. H. Freeman and Company Standardizing a Normal sample distribution (1 of 2) When the sampling distribution is Normal, we can standardize the value of a sample mean x̅ to obtain a z- score. This z-score can then be used to find areas under the sampling distribution from Table B. Copyright © 2018 W. H. Freeman and Company Standardizing a Normal sample distribution (2 of 2) Here, we work with the sampling distribution, and σ /√n is its standard deviation (indicative of spread). Remember that σ is the standard deviation of the original population. Copyright © 2018 W. H. Freeman and Company  Standardization example (1 of 2) Hypokalemia is diagnosed when blood potassium levels are low, below 3.5 mEq/dL. Let’s assume that we know a patient whose measured potassium levels vary daily according to N(µ = 3.8, σ = 0.2). (Note that this mean is high enough that the patient is not considered to be hypokalemic.) If only one measurement is made, what is the probability that this patient will be misdiagnosed hypokalemic? P(z < 1.5) = 0.0668 ≈ 7% Copyright © 2018 W. H. Freeman and Company  Standardization example (2 of 2) If instead measurements are taken on four separate days, what is the probability of such a misdiagnosis? Note: This calculation demonstrates that an average of 4 measurements is more likely to be closer to the true average than individual measurements. Copyright © 2018 W. H. Freeman and Company  The central limit theorem Central limit theorem: When randomly sampling from any population with mean m and standard deviation σ, when n is large enough, the sampling distribution of is approximately Normal: N(µ, σ /√n). The larger the sample size n, the better the approximation of Normality. This is very useful in inference: Many statistical tests assume Normality for the sampling distribution. The central limit theorem tells us that, if the sample size is large enough, we can safely make this assumption even if the raw data appear non-Normal. Copyright © 2018 W. H. Freeman and Company  How large a sample size? It depends on the population distribution. More observations are required if the population distribution is far from Normal. – A sample size of 25 or more is generally enough to obtain a Normal sampling distribution from a skewed population, even with mild outliers in the sample. – A sample size of 40 or more will typically be good enough to overcome an extremely skewed population and mild (but not extreme) outliers in the sample. In many cases, n = 25 isn’t a huge sample. Thus, even for strange population distributions we can assume a Normal sampling distribution of the sample mean, and work with it to solve problems. Copyright © 2018 W. H. Freeman and Company  When the population is skewed Even though the population (a) is strongly skewed, the sampling distribution of when (d) is approximately Normal, as expected from the central limit theorem. Copyright © 2018 W. H. Freeman and Company  Is the population Normal? Sometimes we are told that a variable has an approximately Normal distribution (e.g., large studies on human height or bone density). Most of the time, we just don’t know. All we have is sample data. – We can summarize the data with a histogram and describe its shape. – If the sample is random, the shape of the histogram should be similar to the shape of the population distribution. – The central limit theorem can help guess whether the sampling distribution should look roughly Normal or not. Copyright © 2018 W. H. Freeman and Company  Central limit theorem examples (1 of 4) a) Angle of big toe deformations in 38 patients: Symmetrical, one small outlier Population likely close to Normal Sampling distribution ~ Normal Copyright © 2018 W. H. Freeman and Company  Central limit theorem examples (2 of 4) b) Histogram of number of fruit per day for 74 adolescent girls Skewed, no outlier Population likely skewed Sampling distribution ~ Normal given large sample size Copyright © 2018 W. H. Freeman and Company  Central limit theorem examples (3 of 4) Atlantic acorn sizes (in cm3) Sample of 28 acorns: Describe the distribution of the sample. What can you assume about the population distribution? What would be the shape of the sampling distribution: – For samples of size 5? – For samples of size 15? – For samples of size 50? Copyright © 2018 W. H. Freeman and Company Central limit theorem examples (4 of 4) 14 12 10 8 Frequency 6 4 2 0 1.5 3 4.5 6 7.5 9 10.5 More Acorn sizes Copyright © 2018 W. H. Freeman and Company  Chapter 12: proportions (1 of 2) A population contains a proportion p of successes. If the population is much larger than the sample, the count X of successes in a Simple Random Sample (SRS) of size n has approximately the binomial distribution B(n, p) with mean µ and standard deviation σ: Copyright © 2018 W. H. Freeman and Company  Chapter 12: proportions (2 of 2) If n is large, and p is not too close to 0 or 1, this binomial distribution can be approximated by the Normal distribution: Copyright © 2018 W. H. Freeman and Company Sampling distribution of a proportion (1 of 2) When randomly sampling from a population with proportion p of successes, the sampling distribution of the sample proportion [“p hat”] has mean and standard deviation: Copyright © 2018 W. H. Freeman and Company Sampling distribution of a proportion (2 of 2) – is an unbiased estimator the population proportion p. – Larger samples usually give closer estimates of the population proportion p. Copyright © 2018 W. H. Freeman and Company  Normal approximation The sampling distribution of p̂ is never exactly Normal. But as the sample size increases, the sampling distribution of p̂ becomes approximately Normal. The Normal approximation is most accurate for any fixed n when p is close to 0.5, and least accurate when p is near 0 or near 1.  When n is large, and p is not too close to 0 or 1, the sampling distribution of p̂ is approximately: Copyright © 2018 W. H. Freeman and Company  Numerical example (1 of 2) The frequency of color blindness (dyschromatopsia) in the Caucasian American male population is about 8%. We wish to take a random sample of size 125 from this population. What is the probability that 10% or more in the sample are color-blind? A sample size of 125 is large enough to use of the Normal approximation (np = 10 and n(1 – p) = 115). Copyright © 2018 W. H. Freeman and Company  Numerical example (2 of 2) Normal approximation for p̂ sampling distribution: – z = (p̂ – p) / σ = (0.10 – 0.08) / 0.024 = 0.824  P(z ≥ 0.82) = 0.2061 from Table B – Or P(p̂ ≥ 0.10) = 1 – NORM.DIST(0.10, 0.08, 0.024, 1) = 0.2023 (Excel) = normalcdf (0.10, 1E99, 0.08, 0.024) = 0.2023 (TI-83) Copyright © 2018 W. H. Freeman and Company  The law of large numbers (1 of 3) Law of large numbers: As the number of randomly drawn observations (n) in a sample increases… the mean of the sample () gets closer and closer to the population mean m (quantitative variable). Copyright © 2018 W. H. Freeman and Company  The law of large numbers (2 of 3) the sample proportion () gets closer and closer to the population proportion p (categorical variable). Copyright © 2018 W. H. Freeman and Company  The law of large numbers (3 of 3) Note: When sampling randomly from a given population: – The law of large numbers describes what would happen if we took samples of increasing size n. – A sampling distribution describes what would happen if we took all possible random samples of a fixed size n. Both are conceptual ideas with many important practical applications. We rely on their known mathematical properties, but we don’t actually build them from data. Copyright © 2018 W. H. Freeman and Company