Qsm 754 Six Sigma Minitab Power Point Slides

transcript

©The National Graduate School of Quality Management v.8 • 1

INTRODUCTION TO MINITAB VERSION 13

•Worksheet Conventions and Menu Structures

•Minitab Interoperability

•Graphic Capabilities•Pareto•Histogram•Box Plot•Scatter Plot

•Statistical Capabilities•Capability Analysis•Hypothesis Test•Contingency Tables•ANOVA•Design of Experiments (DOE)

Minitab Training Agenda

Worksheet Format and Structure

Session Window

Worksheet Data Window

Menu Bar

Tool Bar

Text Column C1-T

(Designated by -T)

Numeric Column C3

(No Additional Designation)

Data Window Column Conventions

Date Column C2-D

(Designated by -D)

Column Names

(Type, Date, Count & Amount

Entered Data for Data Rows 1 through 4

Data Entry Arrow

Data Rows

Other Data Window Conventions

Menu Bar - Menu Conventions

Hot Key Available (Ctrl-S)

Submenu Available (… at the end of

selection)

Menu Bar - File Menu

Key Functions

•Worksheet File ManagementSavePrintData Import

Menu Bar - Edit Menu

Key Functions

•Worksheet File EditsSelectDeleteCopyPasteDynamic Links

Menu Bar - Manip Menu

Key Functions

•Data ManipulationSubset/SplitSortRankRow Data ManipulationColumn Data Manipulation

Menu Bar - Calc Menu

Key Functions

•Calculation CapabilitiesColumn CalculationsColumn/Row StatisticsData StandardizationData ExtractionData Generation

Menu Bar - Stat Menu

Key Functions

•Advanced Statistical Tools and GraphsHypothesis TestsRegressionDesign of ExperimentsControl ChartsReliability Testing

Menu Bar - Graph Menu

Key Functions

•Data Plotting CapabilitiesScatter PlotTrend PlotBox PlotContour/3 D plottingDot PlotsProbability PlotsStem & Leaf Plots

Menu Bar - Data Window Editor Menu

Key Functions

•Advanced Edit and Display OptionsData BrushingColumn SettingsColumn Insertion/MovesCell InsertionWorksheet Settings

Note: The Editor Selection is Context Sensitive. Menu selections will vary for:

•Data Window•Graph•Session Window

Depending on which is selected.

Menu Bar - Session Window Editor Menu

Key Functions

•Advanced Edit and Display OptionsFont Connectivity Settings

Menu Bar - Graph Window Editor Menu

Key Functions

•Advanced Edit and Display OptionsBrushing Graph Manipulation

ColorsOrientationFont

Menu Bar - Window Menu

Key Functions

•Advanced Window Display OptionsWindow Management/Display Toolbar Manipulation/Display

Menu Bar - Help Menu

Key Functions

•Help and TutorialsSubject SearchesStatguide Multiple TutorialsMinitab on the Web

MINITAB INTEROPERABILITY

Minitab Interoperability

Minitab

PowerPoint

Starting with Excel...

Load file “Sample 1” in Excel….

The data is now loaded into Excel….

Highlight and Copy the Data….

Move to Minitab...

Open Minitab and select the column you want to paste the data into….

Move to Minitab...

Select Paste from the menu and the data will be inserted into the Minitab Worksheet….

Use Minitab to do the Analysis...

Lets say that we would like to test correlation between the Predicted Workload and the actual workload….

•Select Stat… Regression…. Fitted Line Plot…..

Minitab is now asking for us to identify the columns with the appropriate date….

•Click in the box for “Response (Y): Note that our options now appear in this box.

•Select “Actual Workload” and hit the select button…..

•This will enter the “Actual Workload” data in the Response (Y) data field...

•Now click in the Predictor (X): box…. Then click on “Predicted Workload” and hit the select button… This will fill in the “Predictor (X):” data field...

•Both data fields should now be filled….

•Select OK...

•Minitab now does the analysis and presents the results...

•Note that in this case there is a graph and an analysis summary in the Session Window…

•Let’s say we want to use both in our PowerPoint presentation….

Transferring the Analysis...

•Let’s take care of the graph first….

•Go to Edit…. Copy Graph...

•Open PowerPoint and select a blank slide….

•Go to Edit…. Paste Special...

•Select “Picture (Enhanced Metafile)… This will give you the best graphics with the least amount of trouble.

•Our Minitab graph is now pasted into the powerpoint presentation…. We can now size and position it accordingly….

•Now we can copy the analysis from the Session window…..

•Highlight the text you want to copy….

•Select Edit….. Copy…..

•Now go back to your powerpoint presentation…..

•Select Edit….. Paste…..

•Well we got our data, but it is a bit large…..

•Reduce the font to 12 and we should be ok…..

Presenting the results....

•Now all we need to do is tune the presentation…..

•Here we position the graph and summary and put in the appropriate takeaway...

•Then we are ready to present….

Graphic Capabilities

Pareto Chart....

•Let’s generate a Pareto Chart from a set of data….

•Go to File… Open Project…. Load the file Pareto.mpj….

•Now let’s generate the Pareto Chart...

Pareto Chart....

•Go to:

•Stat…

•Quality Tools…

•Pareto Chart….

Pareto Chart....

Fill out the screen as follows:

•Our data is already summarized so we will use the Chart Defects table...

•Labels in “Category”…

•Frequencies in “Quantity”….

•Add title and hit OK..

Pareto Chart....

Minitab now completes our pareto for us ready to be copied and pasted into your PowerPoint presentation….

Histogram....

•Let’s generate a Histogram from a set of data….

•Go to File… Open Project…. Load the file 2_Correlation.mpj….

•Now let’s generate the Histogram of the GPA results...

Histogram....

•Go to:

•Graph…

•Histogram…

Histogram....

•Select GPA for our X value Graph Variable

•Hit OK…..

Histogram....

Minitab now completes our histogram for us ready to be copied and pasted into your PowerPoint presentation….

This data does not look like it is very normal….

Let’s use Minitab to test this distribution for normality…...

Histogram....

•Go to:

•Stat…

•Basic Statistics…

•Display Descriptive Statistics….

Histogram....

•Select GPA for our Variable….

•Select Graphs…..

Histogram....

•Select Graphical Summary….

•Select OK…..

•Select OK again on the next screen...

Histogram....

Note that now we not only have our Histogram but a number of other descriptive statistics as well….

This is a great summary slide...

As for the normality question, note that our P value of .038 rejects the null hypothesis (P<.05). So, we conclude with 95% confidence that the data is not normal…..

Histogram....

•Let’s look at another “Histogram” tool we can use to evaluate and present data….

•Go to File… Open Project…. Load the file overfill.mpj….

Histogram....

•Go to:

•Graph…

•Marginal Plot…

Histogram....

•Select filler 1 for the Y Variable….

•Select head for the X Variable

•Select OK…..

Histogram....

Note that now we not only have our Histogram but a dot plot of each head data as well...

Note that head number 6 seems to be the source of the high readings…..

This type of Histogram is called a “Marginal Plot”..

Boxplot....

•Let’s look at the same data using a Boxplot….

Boxplot....

•Go to:

•Stat…

•Display Descriptive Statistics...

Boxplot....

•Select “filler 1” for our Variable….

•Select Graphs…..

Boxplot....

•Select Boxplot of data….

•Select OK…..

•Select OK again on the next screen...

Boxplot....

We now have our Boxplot of the data...

Boxplot....

•There is another way we can use Boxplots to view the data...

•Go to:

•Graph…

•Boxplot...

Boxplot....

•Select “filler 1” for our Y Variable….

•Select “head” for our X Variable….

•Select OK…..

Boxplot....

Note that now we now have a box plot broken out by each of the various heads..

Note that head number 6 again seems to be the source of the high readings…..

Scatter plot....

•Let’s look at data using a Scatterplot….

•Go to File… Open Project…. Load the file 2_Correlation.mpj….

•Now let’s generate the Scatterplot of the GPA results against our Math and Verbal scores...

Scatter plot....

•Go to:

•Graph…

•Plot...

Scatter Plot....

•Select GPA for our Y Variable….

•Select Math and Verbal for our X Variables…..

•Select OK when done...

Scatter plot....

We now have two Scatter plots of the data stacked on top of each other…

We can display this better by tiling the graphs….

Scatter plot....

To do this:

•Go to Window…

•Tile...

Scatter plot....

Now we can see both Scatter plots of the data…

Scatter plot....

•There is another way we can generate these scatter plots….

•Go to:

•Graph…

•Matrix Plot...

Scatter Plot....

•Click in the “Graph variables” block

•Highlight all three available data sets…

•Click on the “Select” button...

•Select OK when done...

Scatter plot....

We now have a series of Scatter plots, each one corresponding to a combination of the data sets available…

Note that there appears to be a strong correlation between Verbal and both Math and GPA data….

Minitab Statistical Tools

PROCESS CAPABILITY ANALYSIS

Let’s do a process capability study….

Open Minitab and load the file Capability.mpj….

SETTING UP THE TEST….

Go to Stat… Quality Tools…. Capability Analysis (Weibull)….

Select “Torque” for our single data column...

Enter a lower spec of 10 and an upper spec of 30. Then select “OK”….

SETTING UP THE TEST….

Note that the data does not fit the normal curve very well...

Note that the Long Term capability (Ppk) is 0.43. This equates to a Z value of 3*0.43=1.29 standard deviations or sigma values.

This equates to an expected defect rate PPM of 147,055.

INTERPRETING THE DATA….

HYPOTHESIS TESTING

•Load the file normality.mpj…..

Setting up the test in Minitab

Checking the Data for Normality….

•It’s important that we check for normality of data samples.

•Let’s see how this works….

•Go to STAT…. Basic Statistics... Normality Test….

Set up the Test

•We will test the “Before” column of data….

•Check Anderson-Darling

•Click OK

Analyzing the Results

•Since the P value is greater than .05 we can assume the “Before” data is normal

•Now repeat the test for the “After” Data (this is left to the student as a learning exercise..)

Checking for equal variance..

•We now want to see if we have equal variances in our samples.

•To perform this test, our data must be “stacked”.

•To accomplish this go to Manip… Stack… Stack Columns….

•Select both of the available columns (Before and After) to stack....

•Type in the location where you want the stacked data…. In this example we will use C4….

•Type in the location where you want the subscripts stored… In this example we will use C3….

•Select OK….

•Now that we have our data stacked, we are ready to test for equal variances.…

•Go to Stat… ANOVA…. Test for equal Variances...

Setting up the test….

•Our response will be the actual receipt performance for the two weeks we are comparing. In this case we had put the stacked data in column C4….•Our factors is the label

column we created when we stacked the data (C3).. •We set our Confidence

Level for the test (95%).

•Then select “OK”.

•Here, we see the 95% confidence intervals for the two populations. Since they overlap, we know that we will fail to reject the null hypothesis.

•The F test results are shown here. We can see from the P-Value of .263 that again we would fail to reject the null hypothesis. Note that the F test assumes normality

•Note that we get a graphical summary of both sets of data as well as the relevant statistics….

Analyzing the data….

•Levene’s test also compares the variance of the two samples and is robust to nonnormal data. Again, the P-Value of .229 indicates that we would fail to reject the null hypothesis.

•Here we have box plot representations of both populations.

Lets test the data with a 2 Sample t Test

- -•Under Stat… Basic Statistics…. We see several of the hypothesis tests which we discussed in class. In this example we will be using a 2 Sample t Test….

•Go to Stat…. Basic Statistics.. 2 Sample t…..

•Since we already have our data stacked, we will load C4 for our samples and C3 for our subscripts.

•Since we have already tested for equal variances, we can check off this box…

•Now select Graphs….

•We see that we have two options for our graphical output. For this small a sample, Boxplots will not be of much value so we select “Dotplots of data” and hit “OK”. Hit OK again on the next screen….

•In the session window we have each population’s statistics calculated for us..

•Note that here we have a P value of .922. We therefore find that the data does not support the conclusion that there is a significant difference between the means of the two populations...

Interpreting the results….

•The dotplot shows how close the datapoints in the two populations fall to each other. The close values of the two population means (indicated by the red bar) also shows little chance that this hypothesis could be rejected by a larger sample

Interpreting the results….

Paired Comparisons

In paired comparisons we are trying to “pair” observations or treatments. An example would be to test automatic blood pressure cuffs and a nurse measuring the blood pressure on the same patient using a manual instrument.

It can also be used in measurement system studies to determine if operators are getting the same mean value across the same set of samples.

Let’s look at an example: 2_Hypothesis_Testing_Shoe_wear.mpj

2_Hypothesis_Testing_Shoe_wear.mpj

In this example we are trying to determine if shoe material “A” wear rate is different from shoe material “B”.

Our data has been collected using ten boys, whom were asked to wear one shoe made from each material.

Ho: Material “A” wear rate = Material “B” wear rateHa: Material “A” wear rate Material “B” wear rate

Paired Comparison

•Go to Stat….

• Paired t…..

Paired Comparison

•Select the samples…

•Go to Graphs….

Paired Comparison

•Select the Boxplot for our graphical output..

•Then select OK..

Paired Comparison

We see how the 95% confidence interval of the mean relates to the value we are testing. In this case, the value falls outside the 95% confidence interval of the data mean. This gives us confirmation that the shoe materials are significantly different.

CONTINGENCY TABLES(CHI SQUARE)

Entering the data….

•Enter the data in a table format. For this example, load the file Contingency Table.mpj...

Let’s set up a contingency table….

•Contingency tables are found under Stat…. Tables… Chi Square Test….

•Select the columns which contain the table. Then select “OK”

Note that you will have the critical population and test statistics displayed in the session window.

•Minitab builds the table for you. Note that our original data is presented and directly below, Minitab calculates the expected values.

•Here, Minitab calculates the Chi Square statistic for each data point and totals the result. The calculated Chi Square statistic for this problem is 30.846.

Performing the Analysis….

ANalysis Of VAriance

Let’s set up the analysis

•Load the file Anova example.mpj…•Stack the data in C4 and place the subscripts in C5

Set up the analysis….

•Select Stat…•ANOVA…•One way…

•Select• C4 Responses• C5 Factors•Then select Graphs….

•Choose boxplots of data...•Then OK

Note that the P value is less than .05that means that we reject the null hypothesis

Analyzing the results….

Let’s Look At Main Effects….

•Choose Stat•ANOVA•Main Effects Plot….

Main Effects

Select•C4 Response•C5 Factors•OK

Analyzing Main Effects..

Liters/Hr 3Liters/Hr 2Liters/Hr 1

Formulation

Main Effects Plot - Data Means for Liters Per H

Formulation 1 Has Lowest Fuel Consumption

DESIGN OF EXPERIMENTS (DOE)

FUNDAMENTALS

First Create an Experimental Design...

•Stat…

•DOE…

•Factorial...

•Create Factorial Design...

First Create an Experimental Design...

Select 2 Level Factorial design with 3 factors

Then go to Display Available Designs….

Bowling Example (continued)

We can now see the available experimental designs…. We will be using the Full (Factorial) for 3 factors and we can see that it will require 8 runs…

Now, select OK and go back to the main screen.

Once at the main screen select Designs...

Select your design….

We will be using the Full (Factorial) and again we can see that it will require 8 runs…

Once at the main screen select Factors...

Fill in the names for your factors….

Then fill in the actual conditions for low (-) or high (+)

Once at the main screen select Options...

Remove the option to Randomize Runs….

Once at the main screen select OK...

Minitab has now designed our experiment for us….

Now, type your Data from each of your experimental treatments into C8.

We are now ready to analyze the results…

•Stat….

•DOE…

•Factorial...

•Analyze Factorial Design...

Highlight your Data column and use Select to place it in the Responses box.

Then, select the Terms Option.

Note that Selected Terms has all of the available choices already selected. We need do nothing further.

Select OK.

Then, at the main screen select Graphs

Select your Effects Plots and reset your Alpha to .05.

Select OK to return to the main screen and then select OK again.

Note that only one effect has a significance greater than 95%.

All the remaining factors and interactions are not statistically significant.

•Another way we can look at the data is to look at the Factorial Plots of the resulting data.

•Go to

•DOE….

•Factorial…

•Factorial Plots….

•Select Main Effects Plot and then Setup…

•Select C8 as your response

•Select “Wristband”, “Ball” and “Lane” as your factors.

•Then select “OK” and OK again on the main screen.

•The magnitude of the vertical displacement indicates the strength of the main effect for that factor. Here we see that the wristband has dramatically more effect than any other factor. We know from our earlier plots that the wristband is the only statistically significant effect @ 95% confidence.

•This plot also shows you the direction of the main effects. We clearly see that the “with” condition is related to the higher level of performance.

•Now lets look at the interactions....

•Go to

•DOE….

•Factorial…

•Factorial Plots…

•Select InteractionPlot and then Setup…..

•Select C8 as your response variable.

•Select “Wristband”, “Ball” and “Lane” as your factors.

•Then select “OK” and OK again on the next screen….

•The more the lines diverge from being parallel, the more the interaction.

•We see that the strongest interaction (still not significant) is between the lane and the ball.

•We know from our earlier analysis that none of these interactions were statistically significant for this experiment…..

Bowling Example (Session Window)

•You can also see that there is zero error

•This is because only 1 run was performed with no replications

•This is where Minitab shows us the Main Effects and Interaction Effects..

•Note that Wristband has the strongest effect followed by the interaction between the Wristband and the Lane...

Qsm 754 Six Sigma Minitab Power Point Slides

Documents