About markbailey

markbailey

I am unable to reproduce your outcome. I notice that you have the following analysis roles: Y: Actual NaCl Flux X: Mixed Ion Rej Group Y: Coating Length Group X: Location This set up results in a single value (rectangle) for each grouping. Are you sure you want a tree map? Perhaps a heat map with the same grouping might suffice. Here is an example: I recommend contacting JMP Technical Support (support@jmp.com) about this issue.

markbailey

Uncheck Category Value as well:

markbailey

Is the study design fully balanced? May you share either the data or the JMP platform (results)?

markbailey

Deselect Category Name:

markbailey

You are correct. JMP is not a spreadsheet. You might think of a JMP data table as a collection of observations (i.e., data set). As such, each change in volume is a new row. So you might have one column to represent the volume added or removed and another to represent the current volume. I think the first column, Solution A Removed in Current Draw, should be a value that is entered, not computed. That might break the cyclical references. Possibly the same for the last column, Solution A Added in Current Draw. Adjust the formulas accordingly.

markbailey

I do not mean to contradict one of my colleagues, but how does one specify the factors, model, and number of runs in order to get 'correct design' without correlated estimates? I added four continuous factors representing concentration, temperature, time, and pressure. I added terms for the second and third power of each factor to the linear model. I used the default number of runs. This example is clearly not one of the correct results: I simulated the response so that I could launch Fit Least Squares platform. The response is not involved in the correlation of the estimates, which is also indicated by VIF > 1 in the parameter estimates report. Here is the design: Here is the report using the default coded levels. VIF around 10 indicates a correlation around 0.9. Here is the same analysis but without coded levels. The coding clearly helps: What is the use of the map of correlations if it does not represent the actual correlations? (I think that it does.)

markbailey

Adding to @Thierry_S's suggestion, what if you isolate the Save As() message (forget the function for now) and make sure that it works on its own?

markbailey

Could it be that the data table file is corrupted? Can you open other data tables without crashing JMP?

markbailey

Is the difference in the variances significant?

markbailey

I would suggest using the Functional Data Explorer to analyze the sound samples and then create functional principle components to be used as new predictors or responses. But FDE is only available in JMP Pro 14 and 15. Perhaps someone else has a solution using JMP Pro 13.

markbailey

JMP Community members are happy to help but be warned that DOE is a broad and deep subject. It is difficult to convey knowledge about it in a few replies. Custom Design is unique to JMP but other software implements 'optimal design' methods including the 'coordinate exchange' algorithm. The older methods were based on the technology at hand: paper and pencil and mathematics. It was based on combinatorics and group theory. It was largely from a 'design-centric' view. A new view developed in the 1970s that was more of a 'model centric' view. It is based on an optimality criterion and numerical solution. There are many criteria that can be used. You mention D-optimal and I-optimal designs. D-optimal designs minimize the confidence region of the estimates of the model parameters. This result is best when you are most interested in inference about the estimates. (That is, are they significantly different from zero?) I-optimal designs minimize the integrated variance of estimates of the response. (That is, the model predictions have the smallest confidence intervals.) So you get to choose the way in which the design is optimal. You also get to specify any linear model, if any terms must be estimated or not, aliasing to be avoided, fixed or random blocking, restricted randomization, and the number of runs. You cannot do any of that with either of the two classical designs that you mentioned.

markbailey

@statman describes the assumption for the linear regression model well. GLMs do not have any such assumption about the errors, so your concept of residual analysis does not carry over to the GLM. We can use the various residuals from the fitted GLM to check for outliers, influential observations, and bias, though. The GLM assumes that the response follows the conditional distribution that you selected (e.g., Poisson, binomial, normal) and maximizes the likelihood of the model parameter estimates. It's different.

markbailey

Update the Time value to 34 after the run is finished but before the regression analysis.

markbailey

Consider an I-optimal custom design.

markbailey

Maybe change the column name used in the Y role and change its function to produce some normalized response (transformed scale)?

markbailey

The Analysis of Variance table tests the 'whole model.' That is, the null hypothesis is that all the parameters (except the constant term) are zero versus the alternative hypothesis that not all the parameters are zero. The test is based on the ratio of variances (MS Model / MS Error). This F ratio is used to obtain the p-value that is reported in the Analysis of Variance table and also below horizontal axis of the Actual by Predicted plot. The Effect Summary reports tests for individual parameter estimates. It compares the parameter estimate to the hypothesized parameter (zero) by difference and then makes a t ratio by dividing by the standard error of the estimate. The t ratio is, therefore, the number of standard errors that the estimate is from the hypothesized value. The t ratio is used to obtain the p-value in the Effects Summary table. Note that while both of these tests are asking a question about the parameter estimates, they are not asking the same question. They are also not using the same method to obtain an answer. So they will not necessarily agree or provide consistent answers. For example, the whole model F test might indicate that you should not reject the null hypothesis while one or more of the t tests might indicate that you should reject the null. Remember that they are not the same hull hypothesis.

markbailey

I thought of using the Left() function, too. Here is my take: Names Default to Here( 1 ); string = "AAA-BB-1234-19-001"; Left( string, Contains( string, "00") - 2 ); I recommend that you select Help > Scripting Index and select Functions > Character. There is a lot of built-in support for manipulating character strings.

markbailey

There is a built-in function for this result:

markbailey

Sorry, but I do not know how to perform one-tailed non-parametric tests with JMP. I checked but I could not find them in the Oneway platform.

markbailey

I do not know for sure the differences between JMP and JMP Pro with regard to support for missing value imputation or handling. I am sure about what I posted previously. You can contact JMP Technical Support (support@jmp.com) for more detailed answers. You might want to see this add-in, too. Note: clearly I did not write the documentation!

markbailey

Fit Model is not a platform. It is a user dialog to specify the launch of many different platforms. You can start here in the JMP Help system and find the platform that you intend to use. It should indicate either in the section to launch it or the section about options how it deals with missing values. The Neural platform in JMP Pro supports 'informative missing' instead of imputation. The platform is documented here. The Gaussian Process platform in JMP Pro allows for categorical predictors but it does not provide either imputation or informative missing, so those observations are excluded.

markbailey

JMP does impute missing categorical values. The methods for imputation might depend on the analysis or model of the data, so JMP does not provide a separate imputation function for the data table. Imputation is performed by the analysis platforms. Which platform do you intend to use?

markbailey

The addition of the interaction term to the linear predictor creates a model hierarchy of terms. We generally recommend that you maintain the hierarchy, so do not remove A or B. I can't help with the limitation of not providing a profiler or a way to save the fitted model. You might check the Wishlist area in this Community to see if anyone else has requested it. If so, vote for it! If not, add a request.

markbailey

If I understand " my hypothesis is that there is no difference for median value between two independent groups " correctly, then the null hypothesis is that there is a difference. This situation is sometimes referred to a equivalence test, although it is really still a hypothesis test. It is just that they usual null hypothesis (i.e., no difference) and alternative hypothesis (i.e., difference) are reversed. The case of deciding about the equivalence of the mean of two groups is based on two one-sided tests (TOST) with t-tests. Perhaps you could apply one-side non-parametric tests in the same way. (The underlying issue is that a hypothesis test is one way. Failing to reject the null hypothesis means that you still assume the null hypothesis. You can't say that the null hypothesis is significant.)

markbailey

No, GLM will not always produce smaller p-values than OLS for the same model.

markbailey

Regarding point 1, the main benefit of the GLM is that you are not restricted to a normal distribution model of the response. If your response is (conditionally) normally distributed, then just use OLS.

markbailey

With regard to point 3, OLS is solving the normal equations directly. The results are based on sums of squares. This leads to the F test for the whole model. The normally distributed parameter estimates leads to t tests. On the other hand, the GLM is using MLE. The results are based on -2LogLikelihood. This leads to chi square tests (likelihood ratio tests). So the estimation and tests use different methods. They are not expected to give identical p-values. Note that these values generally agree if not exactly.

markbailey

Also, the estimation procedure differs. The least squares solution is a closed form that is computed directly. The GLM solution is an iterative method for maximum likelihood estimates. Here is the same model estimated first with Fit Least Squares platform: Here is the same model estimated with GLM using the normal distribution for the response with the identity link: The parameterization is the same for the categorical predicts age and sex but the estimation routines differ. The former is based on sums of squares with t and F tests, and the latter is based on likelihood with chi square ratio tests. The two models produce identical predictions of the response:

markbailey

The models might differ by the parameterization of the categorical factors. Please see these details.

markbailey

The GLM is a generalization of the Ordinary Least Squares Regression. They should be essentially equivalent if your case meets the assumptions of ordinary regression. Otherwise, use the GLM with a different error distribution and link function. If you get different p-values, then there is a difference in the model specifications.

Name	Mark Bailey
Location	New Jersey
Date Registered	‎06-23-2011 08:01 AM
Date Last Visited	4 hours ago
Total Messages Posted	3,095
Total Tags	11
Total Kudos Received	2729

Online Status	Online
Date Last Visited	4 hours ago

Subject	Views	Posted
When is a column not a column? jmper-cable	993	Feb 29, 2020 9:58:06 AM
Announcing a new course! JMP Pro Software: Analyzing Patterns in Text Data jmper-cable	2247	Jul 24, 2018 6:40:32 AM
Which of my factors are important? jmper-cable	2893	Sep 20, 2017 6:14:39 AM
New shapes in JMP: Microtitre plates jmp-blog	3343	Mar 27, 2013 10:02:00 AM
Celebrating Statisticians: J. Stuart Hunter jmp-blog	10727	Mar 4, 2013 11:20:06 AM
What’s new in JMP training? Categorical data analysis! jmp-blog	3113	Mar 1, 2012 9:10:41 AM
First Come, First Serve jmp-blog	1990	Mar 18, 2009 10:33:34 AM
What Good Are Error Bars? jmp-blog	18030	Sep 19, 2008 3:17:00 PM
A Trainer's Perspective on the JMP Discovery Conference jmp-blog	1415	Jun 26, 2008 4:42:01 PM
New Edition of the JMP Scripting Course jmp-blog	1320	Dec 4, 2007 7:55:00 PM