Presentation is loading. Please wait.

Presentation is loading. Please wait.

1/29/2016ENGM 720: Statistical Process Control1 ENGM 720 - Lecture 04 Comparison of Means, Confidence Intervals (CIs), & Operating Characteristic (OC)

Similar presentations


Presentation on theme: "1/29/2016ENGM 720: Statistical Process Control1 ENGM 720 - Lecture 04 Comparison of Means, Confidence Intervals (CIs), & Operating Characteristic (OC)"— Presentation transcript:

1 1/29/2016ENGM 720: Statistical Process Control1 ENGM 720 - Lecture 04 Comparison of Means, Confidence Intervals (CIs), & Operating Characteristic (OC) Curves

2 1/29/2016 ENGM 720: Statistical Process Control 2 Assignment: Reading: Chapter 4 Finish reading through 4.3.4 Begin reading 4.4 through 4.4.3 Chapter 8 Begin reading 8 through 8.3 Assignments: Obtain the Hypothesis Test (Chart &) Tables – Materials Page Obtain the Exam Tables DRAFT – Materials Page Verify accuracy as you work assignments Access New Assignment and Previous Assignment Solutions: Download Assignment 2 Solutions Download Assignment 3 Instructions

3 1/29/2016 ENGM 720: Statistical Process Control 3 Hypothesis Tests An Hypothesis is a guess about a situation that can be tested, and the test outcome can be either true or false. The Null Hypothesis has a symbol H 0, and is always the default situation that must be proven unlikely beyond a reasonable doubt. The Alternative Hypothesis is denoted by the symbol H A and can be thought of as the opposite of the Null Hypothesis - it can also be either true or false, but it is always false when H 0 is true and vice-versa.

4 1/29/2016 ENGM 720: Statistical Process Control 4 Hypothesis Testing Errors Type I Errors occur when a test statistic leads us to reject the Null Hypothesis when the Null Hypothesis is true in reality. The chance of making a Type I Error is estimated by the parameter  (or level of significance), which quantifies the reasonable doubt. Type II Errors occur when a test statistic leads us to fail to reject the Null Hypothesis when the Null Hypothesis is actually false in reality. The probability of making a Type II Error is estimated by the parameter .

5 1/29/2016 ENGM 720: Statistical Process Control 5 22 22  θ0θ0 θAθA One-Sided Test Statistic < Rejection Criterion H 0 : θ A ≥ θ 0 H A : θ A < θ 0 Types of Hypothesis Tests Hypothesis Tests & Rejection Criteria DmDm DmDm Two-Sided Test Statistic < -½ Rejection Criterion or Statistic > +½ Rejection Criterion H 0 : -θ 0 ≤ θ A ≤ +θ 0 H A : θ A < -θ 0 or + θ 0 < θ A  One-Sided Test Statistic > Rejection Criterion H 0 : θ A ≤ θ 0 H A : θ A > θ 0 θAθA +θ0+θ0 -θ0-θ0 θAθA θAθA θ0θ0 DmDm H A : M A is lower than M 0 H A : M A is higher than M 0 H A : M A is different than M 0 H 0 : M A is not different than M 0 H 0 : M A is not better than M 0

6 1/29/2016 ENGM 720: Statistical Process Control 6 Hypothesis Testing Steps 1. State the null hypothesis (H 0 ) from one of the alternatives: that the test statistic    ,    ≥  , or    ≤  . 2. Choose the alternative hypothesis (H A ) from the alternatives:      ,      , or      . (Respective to above!) 3. Choose a significance level of the test ( . 4. Select the appropriate test statistic and establish a critical region. (If the decision is to be based on a P-value, it is not necessary to have a critical region) 5. Compute the value of the test statistic (  ) from the sample data. 6. Decision: Reject H 0 if the test statistic has a value in the critical region (or if the computed P-value is less than or equal to the desired significance level  ); otherwise, do not reject H 0.

7 1/29/2016 ENGM 720: Statistical Process Control 7 Testing Example Single Sample, Two-Sided t-Test: H 0 : µ = µ 0 versus H A : µ  µ 0 Test Statistic: Critical Region: reject H 0 if |t| > t  /2,n-1 P-Value: 2 P(X  |t|), where the random variable x has a t-distribution with n _ 1 degrees of freedom

8 1/29/2016 ENGM 720: Statistical Process Control 8 Hypothesis Testing H 0 : μ = μ 0 versus H A : μ  μ 0 t n-1 distribution 0 -|t| |t| P-value = P(X  -|t|) + P(X  |t|)

9 1/29/2016 ENGM 720: Statistical Process Control 9 Hypothesis Testing Significance Level of a Hypothesis Test: A hypothesis test with a significance level or size  rejects the null hypothesis H 0 if a p-value smaller than  is obtained, and accepts the null hypothesis H 0 if a p-value larger than  is obtained. In this case, the probability of a Type I error (the probability of rejecting the null hypothesis when it is true) is equal to . True Situation Test Conclusion CORRECT Type I Error (  ) H 0 is False Type II Error (  ) CORRECTH 0 is True H 0 is FalseH 0 is True

10 1/29/2016 ENGM 720: Statistical Process Control 10 Hypothesis Testing P-Value: One way to think of the P-value for a particular H 0 is: given the observed data set, what is the probability of obtaining this data set or worse when the null hypothesis is true. A “worse” data set is one which is less similar to the distribution for the null hypothesis. H 0 not plausible H 0 plausible Intermediate area 0 0.01 0.10 1 P-Value

11 1/29/2016 ENGM 720: Statistical Process Control 11 Statistics and Sampling Objective of statistical inference: Draw conclusions/make decisions about a population based on a sample selected from the population Random sample – a sample, x 1, x 2, …, x n, selected so that observations are independently and identically distributed (iid). Statistic – function of the sample data Quantities computed from observations in sample and used to make statistical inferences e.g. measures central tendency

12 1/29/2016 ENGM 720: Statistical Process Control 12 Sampling Distribution Sampling Distribution – Probability distribution of a statistic If we know the distribution of the population from which sample was taken, we can often determine the distribution of various statistics computed from a sample, ex: When the CLT applies, the distribution is Normal When sampling for defective units in a large population, use the Binomial distribution When working with the sum of squared Normal distributions, use the  2 -distribution

13 1/29/2016 ENGM 720: Statistical Process Control 13 e.g. Sampling Distribution of the Mean from the Normal Distribution Take a random sample, x 1, x 2, …, x n, from a normal population with mean μ and standard deviation σ, i.e., Compute the sample average x Then x will be normally distributed with mean μ and standard deviation: that is:

14 1/29/2016 ENGM 720: Statistical Process Control 14 Ex. Sampling Distribution of x When a process is operating properly, the mean density of a liquid is 10 with standard deviation 5. Five observations are taken and the average density is 15. What is the distribution of the sample average? r.v. x = density of liquid Ans: since the samples come from a normal distribution, and are added together in the process of computing the mean:

15 1/29/2016 ENGM 720: Statistical Process Control 15 Ex. Sampling Distribution of x (cont'd) What is the probability the sample average is greater than 15? Would you conclude the process is operating properly?

16 1/29/2016 ENGM 720: Statistical Process Control 16

17 1/29/2016 ENGM 720: Statistical Process Control 17 Ex. Sampling Distribution of x (cont'd) What is the probability the sample average is greater than 15? Would you conclude the process is operating properly?

18 1/29/2016 ENGM 720: Statistical Process Control 18 Comparison of Means The first types of comparison are those that compare the location of two distributions. To do this: Compare the difference in the mean values for the two distributions, and check to see if the magnitude of their difference is sufficiently large relative to the amount of variation in the distributions Which type of test statistic we use depends on what is known about the process(es), and how efficient we can be with our collected data Definitely Different Probably Different Probably NOT Different Definitely NOT Different

19 1/29/2016 ENGM 720: Statistical Process Control 19 Situation I: Means Test, Both σ 0 and μ 0 Known Used with: an existing process with good deal of data showing the variation and location are stable Procedure: use the the z-statistic to compare sample mean with population mean  0

20 1/29/2016 ENGM 720: Statistical Process Control 20 Situation II: Means Test σ (s) Known and μ (s) Unknown Used when: the means from two existing processes may differ, but the variation of the two processes is stable, so we can estimate the population variances pretty closely. Procedure: use the the z-statistic to compare both sample means

21 1/29/2016 ENGM 720: Statistical Process Control 21 Situation III: Means Test Unknown σ (s) and Known μ 0 Used when: have good control over the center of the distribution, but the variation changed from time to time Procedure: use the the t-statistic to compare both sample means v = n – 1 degrees of freedom

22 1/29/2016 ENGM 720: Statistical Process Control 22 Situation IV: Means Test Unknown σ (s) and μ (s), Similar s 2 Used when: logical case for similar variances, but no real "history" with either process distribution (means & variances) Procedure: use the the t-statistic to compare using pooled S, v = n 1 + n 2 – 2 degrees of freedom

23 1/29/2016 ENGM 720: Statistical Process Control 23 Situation V: Means Test Unknown σ(s) and μ(s), Dissimilar s 2 Used when: worst case data efficiency - no real "history" with either process distribution (means & variances) Procedure: use the the t-statistic to compare, degrees of freedom given by:

24 1/29/2016 ENGM 720: Statistical Process Control 24 Situation VI: Means Test Paired but Unknown σ (s) Used when: exact same sample work piece could be run through both processes, eliminating material variation Procedure: define variable (d) for the difference in test value pairs (d i = x 1i - x 2i ) observed on i th sample, v = n - 1 dof

25 1/29/2016 ENGM 720: Statistical Process Control 25 Table for Means Comparisons Decision on which test to use is based on answering (at least some of) the following: Do we know the population variance (σ 2 ) or should we estimate it by the sample variance (s 2 )? Do we know the theoretical mean (μ), or should we estimate it by the sample mean ( y ) ? Do we know if the samples have equal-variance (σ 1 2 = σ 2 2 )? Have we conducted a paired comparison? What are we trying to decide (alternate hypothesis)?

26 1/29/2016 ENGM 720: Statistical Process Control 26 Table for Means Comparisons These questions tell us: What sampling distribution to use What test statistic(s) to use What criteria to use How to construct the confidence interval Six major test statistics for mean comparisons Two sampling distributions Six confidence intervals Twelve alternate hypotheses

27 1/29/2016 ENGM 720: Statistical Process Control 27 Ex. Surface Roughness Surface roughness is normally distributed with mean 125 and std dev of 5. The specification is 125 ± 11.65 and we have calculated that 98% of parts are within specs during usual production. This has been the case for a long time. My supplier of these parts has sent me a large shipment. I take a random sample of 10 parts. The sample average roughness is 134 which is within specifications. Test the hypothesis that the lot roughness is higher than specifications at  = 0.05.

28 1/29/2016 ENGM 720: Statistical Process Control 28 Check the hypothesis that the sample of size 10, and with an average of 134 comes from a population with mean 125 and standard deviation of 5. One-Sided Test H 0 :  ≤  0 H A :  >  0 Test Statistic: = Critical Value: Z = 1.645 Should I reject H 0 ? Yes! Since 5.69 > 1.645, it is likely that it exceeds the roughness. e.g. Surface Roughness Cont'd AlphaOne-sidedTwo-sided Level (α)zz 0.11.281551.64485 0.051.644851.95996

29 1/29/2016 ENGM 720: Statistical Process Control 29 ex. cont'd draw the distributions for the surface roughness and sample average 134

30 1/29/2016 ENGM 720: Statistical Process Control 30 e.g. Surface Roughness Cont'd Find the probability that the sample of size 10, and with an average of 134 does not come from a population with mean 125 and standard deviation of 5. = Should I accept this shipment?

31 1/29/2016 ENGM 720: Statistical Process Control 31 e.g. Surface Roughness Cont'd For future shipments, suggest good cutoff values for the sample average (i.e., accept shipment if average of 10 observations is between what and what)? We know that encompasses over 99% of the probability mass of the distribution for x

32 1/29/2016 ENGM 720: Statistical Process Control 32 Operating Characteristic (OC) Curve Relates the size of the test difference to Type II Error (  ) for a given risk of Type I Error (  ) Designing a test involves a trade-off in sample size versus the power of the test to detect a difference The greater the difference in means (  ), the smaller the chance of Type II Error (  ) for a given sample size and  As the sample size increases, the chance of Type II Error (  ) decreases for a specified  and given difference in means (  ).

33 1/29/2016 ENGM 720: Statistical Process Control 33 Operating Characteristic Curve

34 1/29/2016 ENGM 720: Statistical Process Control 34 Agree on acceptable  Need to have an OC curve for the correct hypothesis test and the correct  level Estimate anticipated  and  to compute d: d = |  1 -  2 | = |  |  Look for where d intersects with desired  (Probability of accepting H 0 ) to estimate the required sample size (n) O.C. Curve Use

35 1/29/2016 ENGM 720: Statistical Process Control 35 OC Curve Example (uses Fig 3-7, p.111) Assume our previous problem had a process std. dev. of 18 (instead of 5), and the same means (125 population & spec, 134 supplier sample). Assume the boss wants  = 0.05 of exceeding either the high or low spec. for such a sample. Probability of what (in English)? Contracting an incapable supplier, based on a bad-luck test outcome Assume supplier needs  = 0.2 Probability of what (in English)? Unfairly being the incapable supplier, based on a bad-luck test outcome What sample size is needed to fit these constraints?

36 1/29/2016 ENGM 720: Statistical Process Control 36 Two-Sided Operating Characteristic Curve,  = 0.05 d = 0.5 β = n = 30

37 1/29/2016 ENGM 720: Statistical Process Control 37 Estimation of Process Parameters In SPC: the probability distribution is used to model a quality characteristic (e.g. dimension of a part, viscosity of a fluid) Therefore: we are interested in making inferences about the parameters of the probability distribution (e.g. mean μ and variance σ 2 ) Since: Values of these parameters are generally not known, so we need to estimate them from sample data

38 1/29/2016 ENGM 720: Statistical Process Control 38 Point Estimate Numerical value, computed from a sample of data, used to estimate a parameter of a distribution Example: Say we take n = 50 measurements of a quality characteristic Sample mean is point estimate of μ i.e. Sample variance is point estimate of σ 2 i.e.

39 1/29/2016 ENGM 720: Statistical Process Control 39 Confidence Intervals A confidence interval for an unknown parameter  is an interval that contains a set of likely values of the parameter. It is associated with a confidence level 1- , which measures the probability that the confidence interval actually contains the unknown parameter. θ

40 1/29/2016 ENGM 720: Statistical Process Control 40 Confidence Interval (C.I.) (Interval Estimate) A C.I. is an interval that, with some probability, includes the true value of the parameter Ex. C.I. of mean μ is L - lower confidence limit U - upper confidence limit (1-  - probability that true value of parameter lies in interval (we pick  ) The interval L  μ  U is called a 100(1-  )% C.I. for the mean

41 1/29/2016 ENGM 720: Statistical Process Control 41 C.I. on the Mean of Normal Distribution with Variance Unknown Suppose, and We don't know the true mean μ or true variance σ 2 A 100(1-  )% C.I. for the unknown (true) mean μ is: - sample mean s - sample standard deviation n - number of observations in sample - value of t distribution

42 1/29/2016 ENGM 720: Statistical Process Control 42 Ex. C.I. on the Mean of Normal Distribution with Variance Unknown Automatic filler deposits liquid in a container. WANT: 95% C.I. on the mean amount (ounces) per container Collect random sample: x 1, x 2, …, x n say n = 10 Compute sample average: Compute sample variance:

43 1/29/2016 ENGM 720: Statistical Process Control 43 Ex. C.I. on Mean cont'd Find the t-distribution value: Look in Table (Appendix IV) Want a 95% C.I. so, 100(1 -  )% = 95%   = 0.05 = degrees of freedom = (n -1) = 9 so …

44 1/29/2016 TM 720: Statistical Process Control 44

45 1/29/2016 ENGM 720: Statistical Process Control 45 Ex. C.I. on Mean cont'd Find the t-distribution value: Look in Table (Appendix IV) Want a 95% C.I. so, 100(1 -  )% = 95%   = 0.05 = degrees of freedom = (n -1) = 9 so … Substitute into C.I. - or -

46 1/29/2016 ENGM 720: Statistical Process Control 46 Interpretation of a 95% C.I. Repeat sampling 10,000 (or many, many) times & obtain C.I.s Each C.I. will have (slightly) different center point and width On average, 95% of the C.I.s will include the true mean

47 1/29/2016 ENGM 720: Statistical Process Control 47 C.I.s on Other Parameters and Quantities Same procedure, different formulas For example, C.I. on Mean (of any distribution) when variance is known Variance of a normal distribution Difference in two means (of any distribution) when variances are known Difference in two means from normal distribution when variances are unknown Ratio of variances of two normal distributions etc.... (See textbook Sections 4.3.1, 4.3.4 to review derivations)

48 1/29/2016 ENGM 720: Statistical Process Control 48 Questions & Issues


Download ppt "1/29/2016ENGM 720: Statistical Process Control1 ENGM 720 - Lecture 04 Comparison of Means, Confidence Intervals (CIs), & Operating Characteristic (OC)"

Similar presentations


Ads by Google