220 likes | 300 Views
Section IX- Analyze. A Simple Linear Model. X. Y. We first looked at this for the regression analysis. Introduction to (DOE) – one factor test Develops a model relating a response to a factor - allows prediction of responses at levels other than where data was collected
E N D
A Simple Linear Model X Y • We first looked at this for the regression analysis. • Introduction to (DOE) – one factor test • Develops a model relating a response to a factor - allows prediction of responses at levels other than where data was collected • quantifies the strength of the model • identifies outliers • Uses terminology involved in simple linear regression • Intercept • Slope • Prediction Equation Y = f(X) • R-squared IX-8
Best Fit Example Draw a “best fit” line through the data Place the values you found for bo and b1 in the following template. y= + X bob1 Where bo = y-intercept b1 = slope = Rise/Run IX-9
Answer How tall do we predict someone would be if their shoe size is 9.5? IX-12
Examples of ‘Correlation’ is a fancy word for how well the model predicts the response from the factors. IX-18
Shoe Size vs. Height Exercise Same answer as slide 5? What is the r² best fit line? IX-18
How Do You Show That You Made a Difference? • Did your project reduce variation or shift the mean? • How do you demonstrate improvements statistically IX-24
Hypothesis Testing The question one needs to ask during a project is whether the changes made as the result of a Green Belt project have made a difference in the process. The question is, “Has the project made a difference, and if so, how confident are we that the difference is statistically significant?” We strive for 95% confidence or better (α=.05). IX-24
Hypothesis • A hypothesis is: • a tentative explanation for an observation that can be tested by further investigation. • taken to be true for the purpose of investigation; an assumption. • Examples: • Average gas mileage differs depending on type A or B gas. • Probability of death in auto accidents differs depending on seat belt usage. • The type of aspirin determines the amount of pain relief. • A Six Sigma Green Belt project produced significant results with regard to mean and/or standard deviation improvement. IX-24
Hypothesis Basics • We set up 2 hypotheses • H0 is called the null hypothesis • what is being tested • must contain an = • assumed to be true • H1 is called the alternate hypothesis • Based on the data we collect, we must decide in favor of either H0 or H1. Which does the evidence support? • can only reject null, or fail to reject null IX-24
n = sample size Z/2= 1.96 for 95% confidence = 2.576 for 99% confidence = estimated standard deviation error level = amount of error allowed (i.e. to estimate within +/-2 standard deviation, the half interval width is 2) Sample Size (variable data) Based on a normal distribution Z values from the Standard Normal table IX-31
Example A test on cigarettes yielded an average nicotine content of 15.6 milligrams and a standard deviation of 2.1 milligrams. If we want to be 99% confident in detecting a 1 milligram change, what sample size would be required? What if we only want to be 95% confident? IX-31
Sample size (attribute data) n = sample size Z/2 = 1.96 for 95% confidence = 2.576 for 99% confidence = average proportion defectives = q (proportion non-defectives) = amount of error allowed Based on a binomial distribution IX-31
Example You have just received a shipment of computer memory chips. Suppose we wanted to have 95% confidence with detecting a change of 1%. Assuming the historical proportion of defectives in no more than 2%, what sample size would be required? IX-31
Confidence interval Up to this point we have talked about point estimators (mean, standard deviation, etc) these represent population parameters. We can also construct an interval that has a predetermined probability of including the true population parameter (). This is a confidence interval. IX-34
Example A test on a random sample of 9 cigarettes yielded an average nicotine content of 15.6 milligrams and a standard deviation of 2.1 milligrams. Construct a 99% confidence interval for the true average nicotine content. What if we only want 95% confidence? What happens to the 99% confidence interval if we had tested 30 cigarettes originally? IX-35
Example You have just received a shipment of computer memory chips. From a sample of 100 chips, you find 6 to be defective. Find a 95% confidence interval for the true but unknown proportion of defective chips in the shipment. What if we want 99% confidence? IX-36
t-Test Example Let’s set the Statapult up and take some data. Set the Statapult up and reduce variation as much as possible. Confirm with >25 shots. Calculate and record average . Choose a condition to test that you expect will make a difference in the mean. Use two inspectors. Record each inspectors answers separately. Fire 10 shots with the new setting. Calculate sample average, X-bar and sample standard deviation from one of the inspectors. Save the other inspector data for a later example. (1) Create a hypothesis test to determine if the changed condition results in a statistical difference. (2) Determine if there is a difference between the two inspectors. IX-45
F-Test Example Let’s set the Statapult up and take some data. Set the Statapult up and reduce variation as much as possible. Confirm with 10 shots. Calculate and record sample standard deviation s, and variance s². Choose a condition to test that you expect will make a difference in the variation. Shoot 10 shots with the new condition and calculate sample standard deviation s, and variance s². Create a hypothesis test to determine if the changed condition results in a statistical difference. IX-48