Haomin Yang School of Public Health Fujian Medical University Content Sampling error and Sampling distribution Central limit theorem Standard error t distribution Point estimation Confidence Interval estimation
2 2 Normal distribution
Any normal distribution can be transformed into
the standard normal distribution.
X X ~ N ( , ) 2 Z
3 Critical value
Z / 2 : Two sided critical value
Z : One sided critical value
Critical value Area of Area of
one tail two tails 1.645 0.05 0.10 1.960 0.025 0.05 2.576 0.005 0.01
Z
4 The relationship between the population and sample
Population (The complete set) inference
sampling Sample (The subset of the population)
Samples are taken from populations to provide estimates
of population parameters. Then we use sample data to make an inference about a population.
5 Sampling error
The difference between statistics from different
samples, as well as the difference between sample statistics and population parameter, is called sampling error. Sampling distribution
a probability distribution of a statistic
obtained from a larger number of samples(with sample size N) drawn from a specific population, usually the mean Central limit theorem For simple random samples of n observations taken from a population with mean equal to μ and standard deviation equal to σ, regardless of the population’s distribution, provided the sample size n is sufficiently large, the distribution of the sample mean will be approximately normal with mean equal to μ and standard deviation equal to the population standard deviation divided by the square root of the sample size. Standard error
The variation of the sample mean, or the
standard deviation of the sample mean, is called the standard error of the mean SE, denoted by X . n Standard error
In practice, the population standard deviation σ
is usually unknown and replaced by the sample standard deviation s approximately.
S X SX n n
10 t distribution
If the population standard deviation σ is unknown
and replaced by the sample standard deviation s, at this time, it is t transformation.
X X Z ╳ t SX S X SX n
11 It was developed by William Sealy Gosset under the pseudonym Student.
12 t distribution t value obeys t distribution with the degree of freedom v. Standard normal distribution
X t ，t ~t ( )， =n 1 SX
14 The properties of t distribution:
The t distribution is a family of curves, each is
determined by a degree of freedom v. The t distribution is bell shaped and symmetric about 0.
15 The properties of t distribution:
As the degree of freedom increases, the peak
of curve is higher and the base is more narrow and the t distribution approaches the standard normal distribution. The t distribution with v=∞ is the standard normal distribution.
16 Area under t distribution curve
t ,
One sided critical value：t ,
P (t t ,v ) ，P (t t ,v )
17 Area under t distribution curve
-t 2, t 2,
Two sided critical value：t 2,
P ( t t 2,v ) P (t t 2,v ) P (t t 2,v ) P (t 2,v t t 2,v ) 1- 18 19 Example
When ν =9 and one side probability α =0.05, how
much is t , ?
When ν =9 and two sides probability α =0.05,
how much is t 2, ?
20 21 When ν =9 and one side probability α =0.05, how much is t , ? t , =1.833
The Difference Between a T Distribution and a Normal Distribution Normal distributions are used when the population distribution is assumed to be normal. The T distribution is similar to the normal distribution, just with fatter tails. Both assume a normally distributed population. T distributions have higher kurtosis than normal distributions. The probability of getting values very far from the mean is larger with a T distribution than a normal distribution. 25 KEY TAKEAWAYS •The T distribution is a continuous probability distribution of the z-score when the estimated standard deviation is used in the denominator rather than the true standard deviation. •The T distribution, like the normal distribution, is bell- shaped and symmetric, but it has heavier tails, which means it tends to produce values that fall far from its mean. •T-tests are used in statistics to estimate significance 26 Exercise
The total area under a t-curve is:
A.10 B.100 C.1/2 D.1 Exercise
A t-curve is symmetrical around:
A. Its standard deviation
B. 100 C. 10 D. 0 Exercise
A t-curve never:
A. Touches the y-axis
B. Touches the vertical axis C. Looks like the standard normal distribution curve D. Touches the horizontal axis example
The CEO of light bulbs manufacturing company claims that
an average light bulb lasts 300 days. A researcher randomly
selects 15 bulbs for testing. The sampled bulbs last an
average of 290 days, with a standard deviation of 50 days. If
the CEO’s claim were true, what is the probability that 15
randomly selected bulbs would have an average life of no
more than 290 days?
Exercise A chemical engineer has the following results for the active ingredient yields from 16 pilot batches processed under a retorting procedure:X = 32 grams/liter, s = 3 Determine the approximate probability for getting a result this rare or rarer if the true mean yield is 30.5 grams/liter Exercise
Suppose that T has the t distribution with n = 10
degrees of freedom. For each of the following, compute the true value using Excel and then compute the normal approximation. Compare the results.
a.ℙ(−0.8 < T < 1.2)
b. The 90th percentile of T . Parameter estimation & confidence interval Statistical analysis
35 Parameter Given a model, the parameters are the numbers that yield the actual distribution.
In the real world often you don’t know the “true” parameters, but you get to observe data. Next up, we will explore how we can use data to estimate the model parameters. Statistics as Estimators
We use sample data compute statistics.
The statistics estimate population values, e.g., X
An estimator is a method for producing a best guess
about a population value. An estimate is a specific value provided by an estimator. We want good estimates. What is a good estimator? What properties should it have? Parameter estimation Using the sample statistics to estimate the population parameter is called parameter estimation. There are two types of estimations for population parameter: Point estimation & Interval estimation
39 Point estimation
Directly use sample statistics to estimate
population parameter. The point estimation of the population mean μ is the sample mean x . The point estimation of the population standard deviation σ is the sample standard deviation s.
40 Point estimation Point estimation represents our best “determination” of the parameter. However, it does not express the uncertainty in the estimation. Point estimation does not consider sampling errors.
41 A best estimator
Unbiased: any sample statistic whose expected
value is equal to the parameter. o Mean of the sampling distribution 𝑥ҧ is equal to 𝜇
Minimum variance: no other sample statistic
based on random samples of size n has a sampling distribution with a smaller variance Methods of Point Estimation
1. Method of Moments
2. Maximum Likelihood 3. Bayesian Let M1, M2,... be independent random variables having a common distribution possessing a mean µ. Then the sample means converge to the distributional mean as the number of observations increase. Interval estimation
An interval estimation of a parameter is an
interval or a range of values used to estimate the parameter. A confidence interval (CI) is a specific interval estimation of a parameter determined by using data obtained from a sample.
47 Confidence interval (CI)
There is a 1-α probability that an interval
containing the population mean. This interval is called a 1-α confidence interval (CI) of the population mean.
48 Confidence interval (CI) 1-α CI denoted by (A, B).
lower confidence limit upper confidence limit
CI is open interval. In practice, 1-α is usually 90%, 95% and 99%.
49 Methods to calculate the confidence interval
Parametric interval estimation
Method of normal distribution: n is large Method of t distribution: n is small Nonparametric interval estimation (bootstrap)
50 Method of normal distribution
2 X ~ N ( , ) n X Z ，Z ~N (0,1) α/2 1-α α/2 X P ( Z 2 Z Z 2 ) 1 -Z 2 0 Z 2
X P(Z 2 Z 2 ) 1 X P(Z 2 X X Z 2 X ) 1
P( X Z 2 X X Z 2 X ) 1
( X Z 2 X , X Z 2 X ) ( X Z 2 S X , X Z 2 S X )
51 Method of normal distribution
When the sample size is large, say n greater than
or equal to 100, not only is the sampling distribution of sample means well approximated by the normal distribution, but the sample standard deviation, s, is a reliable estimate of the population standard deviation, σ, which is usually also not known.
52 example
The average height of 120 boys who were seven
year-old was 123.62cm and the standard deviation was 4.75cm. Calculate the 90% confidence interval of the population mean of height for 7 year-old boys.
53 answer
n 120, X 123.62, S 4.75, Z 2 Z 0.10 2 1.645
( X Z 2 S X , X Z 2 S X ) 4.75 4.75 (123.62 1.645 ,123.62 1.645 ) 120 120 (122.91,124.33)
54 Method of t distribution
However, for small sample (n<100), s is far from σ,
so the method of normal distribution cannot be used.
Because σ is unknown for small sample, we cannot
do the Z transformation for sample means.
55 Method of t distribution
2 X ~ N ( , ) n X t ，t ~t (n 1) SX -t t P(t 2,v t t 2,v ) 1 2, 2,
X ( X t 2,v S X , X t 2,v S X ) P(t 2,v t 2,v ) 1 SX P(t 2,v S X X t 2,v S X ) 1 P( X t 2,v S X X t 2,v S X ) 1
56 Example
Randomly select 25 healthy adult males with
average pulse equal to 73.6 beats/min and standard deviation equal to 6.5 beats/min. calculate the 95% confidence interval of the population mean of pulse for healthy adult males.
57 Answer
n 25, X 73.6, S 6.5, t 2,v t0.05 2,24 2.064
( X t 2,v S X , X t 2,v S X ) 6.5 6.5 (76.3 2.064 ,76.3 2.064 ) 25 25 (73.62,78.98) 58 How to interpret the 95% CI
We can increase the confidence (accuracy)
that the interval covers the population mean by increasing the confidence level 1-α. The effect of increasing the confidence level will increase the width of the confidence interval. (decrease the precision)
59 How to interpret the 95% CI
To improve the precision of the confidence
interval, we can decrease the confidence level. To improve the accuracy and precision at the same time, the only one method is to increase the sample size.
60 How to interpret the 95% CI
It is incorrect to interpret a 95% CI by saying that
“there is a 95% probability that the population mean lies within the CI”. Either the population mean is in the interval or it is not in the interval. It is deterministic rather than probabilistic.
61 How to interpret the 95% CI
A 95% CI indicates that if we repeatedly draw 100
independent, random samples from the same population and calculate 100 95% confidence intervals of population mean using these samples, theoretically, there are 95 confidence intervals will contain the population mean, and 5 will not.
62 How to interpret the 95% CI
63 How to interpret the 95% CI However, most statisticians often describe confidence intervals in this way: the value of 0.95 is really the probability that the limits calculated from a random sample will include the population value. For 95% of the calculated confidence intervals it will be true to say that the population mean, μ, lies within this interval.
64 How to interpret the 95% CI
The problem is with a single study we just do not
know which one of these 100 intervals we will obtain and hence we will not know if it includes μ. So we usually interpret a confidence interval as the range of values within which we are 95% confident that the true population mean lies.
65 Exercise What is meant by the term “90% confident” when constructing a confidence interval for a mean? A. If we took repeated samples, approximately 90% of the samples would produce the same confidence interval. B. If we took repeated samples, approximately 90% of the confidence intervals calculated from those samples would contain the sample mean. C. If we took repeated samples, approximately 90% of the confidence intervals calculated from those samples would contain the true value of the population mean. D. If we took repeated samples, the sample mean would equal the population mean in approximately 90% of the samples Exercise Among various ethnic groups, the standard deviation of heights is known to be approximately three inches. We wish to construct a 95% confidence interval for the mean height of male Swedes. Forty-eight male Swedes are surveyed. The sample mean is 71 inches. The sample standard deviation is 2.8 inches. 1.x¯=________ 2.σ =________ 3.n=________ 2.In words, define the random variables X and X¯ .
3.Which distribution should you use for this problem? Explain your choice. 4.Construct a 95% confidence interval for the population mean height of male Swedes. 1.State the confidence interval.. 5.What will happen to the level of confidence obtained if 1,000 male Swedes are surveyed instead of 48? Why? 71 3 48 X is the height of a Swedish male, and x is the mean height from a sample of 48 Swedish males. Normal. We know the standard deviation for the population, and the sample size is greater than 30. CI: (70.15, 71.85) The confidence interval will decrease in size, because the sample size increased. Recall, when all factors remain unchanged, an increase in sample size decreases variability. Thus, we do not need as large an interval to capture the true population mean. Exercise A pharmaceutical company makes tranquilizers. It is assumed that the distribution for the length of time they last is approximately normal. Researchers in a hospital used the drug on a random sample of nine patients. The effective period of the tranquilizer for each patient (in hours) was as follows: 2.7; 2.8; 3.0; 2.3; 2.3; 2.2; 2.8; 2.1; and 2.4. 1.x¯= __________ 2.sx= __________ 3.n= __________ 4.n–1= __________ 2.Define the random variable Xin words. 3.Define the random variable X¯in words. 4.Which distribution should you use for this problem? Explain your choice. 5.Construct a 95% confidence interval for the population mean length of time. 1.State the confidence interval. 2.Sketch the graph. 3.Calculate the error bound. 6.What does it mean to be “95% confident” in this problem? 1.x¯=2.51 2.sx=0.318 3.n=9 4.n−1=8 2.the effective length of time for a tranquilizer 3.the mean effective length of time of tranquilizers from a sample of nine patients 4.We need to use a Student’s-t distribution, because we do not know the population standard deviation. 1.CI:(2.27,2.76) 2.Check student's solution. 3.EBM:0.25 6.If we were to sample many groups of nine patients, 95% of the samples would contain the true population mean length of time. 72 Exercise The average height of young adult males has a normal distribution with standard deviation of 2.5 inches. You want to estimate the mean height of students at your college or university to within one inch with 93% confidence. How many male students must you measure? EBM= error bound for the mean
