[JMP SEA] JMP on Tuesday Session #4 : Many Means Tutorial
Created:
JMP on Tuesday Session #4, 12th May, 2020
Presenter : Bass Masri (@Bass_Masri )
Topic : Beginner's Tutorial
Thank you for your questions during the session. We're sharing the answers here for everyone and for us to improve our JMP skills together.
Recorded video of the session
Q1: How do you read the chart on the right in the Ordered Differences Report?
A1: The way to read the Ordered Differences Report is to look at each row and examine the p-value.
- Using the Companies.jmp sample data set, the first row looks at the difference between big and small size companies profits.
- The null hypothesis is that there is no significant difference between big and small size companies profits.
- Given the p-value is less than 0.05 we reject the null hypothesis.
- And conclude that there is evidence to suggest a significant difference in profits between big and small companies.
- The difference is around $575M profits and we can be 95% sure the difference could be as low as $190M and as high as $959M
You follow a similar process for the second row medium and small companies – we can conclude there is also a significant difference in profits of around $413M between medium and small companies.
For the third row big and medium companies the p -value is greater than 0.05 therefore there is no evidence to suggest that the profits differ for big and medium companies.
I highly recommend the Learning Library for more specific information on hypothesis testing and the one way anova test.
https://www.jmp.com/en_au/learning-library/basic-inference-proportions-and-means.html
Q2 : How do I link the histogram and blue lines to the p-values?
A2 : The [histogram] bar represent the size of the difference and the blue line is meant to indicate direction (positive or negative difference).
The larger the bar (difference) the smaller the p value. When the p-value is less than 0.05 then we consider it a statistically significant difference.
Note that you can select the Help Icon from the Tools Menu and drag it to any Report Output you like and JMP®15 will take you directly to the JMP online help.
Q3: What was the DOE behind the drug dataset? Is it a completely randomised design (CRD)?
A3: The data set Drugs.jmp was a sample data set that you can find under the
Help > Tutorials > Many Means Tutorial.
The information about the data set is limited to that it is made up of one continuous response variable and one categorical variable with three levels/types.
I highly recommend the JMP Community to read discussions and blog posts and particularly the File Exchange to search for sample data files.
Q4 : What is the criterion for choosing Fit Y by X and Fit Model?
A4 : The Fit Y by X platform models the relationship of two variables at a time (Bivariate or one response and one predictor). Whereas the Fit Model Platform can model multiple variables to a response variable.
Fit Y by X Overview
https://www.jmp.com/support/help/en/15.1/#page/jmp/introduction-to-fit-y-by-x.shtml#
Fit Model Overview
https://www.jmp.com/support/help/en/15.1/#page/jmp/overview-of-the-fit-model-platform.shtml#
Q4: Can you use the ANOVA when raw data is not the same between groups?
A4: Yes, you can use the ANOVA test if the variation is different within groups. You may need to use the Unequal Variances test which provides a Welch test on the means and assumes the variances are different.
Q5 : May I ask what is the difference between 1. Each Pair, Student T and 2. All Pairs, Tukey HSD?
A5 : The Each Pair, Student T tests for all possible individual combinations, there is no adjustment for multiple tests. The All Pairs Tukey Honest Significant Different Tests protects the overall error rate and considers the sample size. Read more about the Compare Means for One Way ANOVA online at the JMP®15 Help Support page.
https://www.jmp.com/support/help/en/15.1/#page/jmp/compare-means.shtml#ww873816
Q6 : How to identify which pairs [groups] to use while using 'compare means'?
A6 : You can easily identify the group or pairs by selecting the circle in the Each Pair Test on the right hand side and this will identify the group in the chart.
Read more about the Compare Means for One Way ANOVA online at the JMP®15 Help Support page.
https://www.jmp.com/support/help/en/15.1/#page/jmp/compare-means.shtml#ww873816
Q7 : Do we need to check assumption of Homogeneity of Variance when using one way ANOVA??
A7 : One-way ANOVA can only be used when investigating a single factor and a single dependent variable. When comparing the means of three or more groups, it can tell us if at least one pair of means is significantly different, but it can’t tell us which pair. Also, it requires that the dependent variable be normally distributed in each of the groups and that the variability within groups is similar across groups.
You can use the ANOVA test if the variation is different within groups. You may need to use the Unequal Variances test which provides a Welch test on the means and assumes the variances are different.
I highly recommend the Statistics Knowledge Portal which is a free online introduction to statistics.
https://www.jmp.com/en_au/statistics-knowledge-portal.html
Q8 : Can I use this to compare the SD of a set of repeat measurements, to find which experimental conditions are poorly reproducible?
A8 : One-way ANOVA is used when investigating a single factor and a single dependent variable. I would highly recommend using a Measurement Systems Analysis which is sometimes known as a Gauge Repeatability and Reproducibility Study to asses the variation (SD) in your repeated measures as well as the variation within and between operators. Measurement Systems Analysis can be found under:
Analyze > Quality and Process > Measurement Systems Analysis and Variability / Attribute Gauge Chart.
I’d highly recommend exploring the Statistical Thinking for Problem Solving e-learning course. The course is broken down into seven practical modules including decision making with data, quality methods and design of experiments. Each module comes with its own certificate of completion, which is especially useful for people that want to add statistical techniques to their skills development plans.
https://www.jmp.com/en_au/online-statistics-course.html
Q9 : How can I select which post-hoc test is suitable for which situation?
A9 : One-way ANOVA can only be used when investigating a single factor and a single dependent variable. When comparing the means of three or more groups, it can tell us if at least one pair of means is significantly different, but it can’t tell us which pair. Under Compare Means there are five main tests that you can use to make decisions about differences between individual combinations… For example the Each Pair, Student T tests for all possible individual combinations, there is no adjustment for multiple tests. The All Pairs Tukey Honest Significant Different Tests protects the overall error rate and considers the sample size.
Read more about the Compare Means Tests for One Way ANOVA online at the JMP®15 Help Support page.
https://www.jmp.com/support/help/en/15.1/#page/jmp/compare-means.shtml#ww873816
Q10 : Can we use anova to know the test results is significant or non significant?
A10 : Yes, the one-way ANOVA is used when investigating a single factor and a single dependent variable. When comparing the means of three or more groups, it can tell us if at least one pair of means is statistically significantly different using the p-value obtained from the F Tests which compares within and between group variation. If you only have two groups consider the Fit Y by X platform and the T Test.
I’d highly recommend exploring the Statistical Thinking for Problem Solving e-learning course. The course is broken down into seven practical modules including decision making with data, quality methods and design of experiments. Each module comes with its own certificate of completion, which is especially useful for people that want to add statistical techniques to their skills development plans.
https://www.jmp.com/en_au/online-statistics-course.html
Q11 : Is there any rules for selecting a perfect post-hoc test in ANOVA?
A11 : I would say there is no rule for selecting a perfect post-hoc test in ANOVA, it depends on your research question, sample size, whether you want to consider overall rates and whether you want to compare means or the max and mins.
I highly recommend reading more about the Compare Means Tests for One Way ANOVA online at the JMP®15 Help Support page.
https://www.jmp.com/support/help/en/15.1/#page/jmp/compare-means.shtml#ww873816
Also, I’d highly recommend exploring the Statistical Thinking for Problem Solving e-learning course. The course is broken down into seven practical modules including decision making with data, quality methods and design of experiments. Each module comes with its own certificate of completion, which is especially useful for people that want to add statistical techniques to their skills development plans.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.