JMP Wish List

現在JMPにおいては、PearsonもSpearmanもそのP値は、（z変換不要の）t分布近似の値である。しかし、Fisher z変換からのP値算出を切に熱望する。特に、最近の生物系の投稿論文では、相関係数PearsonもSpearmanも、よく見かける。その場合、Fisher z変換からのP値算出が圧倒的に多い。よろしくお願いしたい。また、信頼区間構成も可能な限りFisher z変換による95%信頼区間構成を望むものである。（なお、本件に関して正確法までは不要とする） ... View more

In an unequal slopes ANCOVA model the individual slopes can be estimated with the Estimates > Custom test or Estimates > Compare slopes option in the output for Fit model. Standard errors and p-values are also given but I miss confidence intervals for the slopes. This should also be possible if additional random effects are added to the model (a mixed model). ... View more

The Profiler is a powerfull tool in JMP and widely used for the analysis and interpretation of Design of Experiment models. To make prediction about future runs, the uncertainity of a model has to be considered. However, JMP is currently limited to confidence intervals, which only allow to make predictions about the range where the true mean falls into. Prediction and tolerance intervals on the other side would extend this so single runs or even for a poportion of a population. Other software tools such as DesignExpert offer these options already for years and I guess many other JMP users would appreacitae such an option in one of the future JMP versions.     ... View more

Add a test to produce p values of superiority tests at a range of distances and output those to a table or graph as shown below.   Consider a trial where a process change is made.  Results are compared with current performance and a t test determines that the process was improved (the mean shifted) with a p-value of 0.05 and a superiority test in JMP 17 indicates that the mean shifted by at least 0.01 with a p value of 0.12, but neither tells a whole story.  To make an informed decision, we would want to know information like: 5% chance the process stayed the same or got worse 90% chance the process improved by at least 0.01 50% chance the process improved by at least 0.006 5% change the process improved by at least 0.011 Here is a post with two potential methods of calculating these values: Compare means test: graph probability by difference in means  ... View more

Can you please add in the ability to lock difference in means, total sample sizes, and power (having to leave one un locked) to all power platforms?  That way you can vary the different locked parameters and determine the effect on the unlocked one?   Thanks!   Steve ... View more

Could you please add Min and max values to the output table?   ... View more

Hi everyone,   I'm using the functional DoE Platform and have to admit that it works really great. However, I miss the possibility to visualize the uncertainty of my models in the profiler. I think a lot of people could benefit from a feature which allows to add confidence intervals or CI's based on a bootstrapping approach. I would really apricate some upvotes for my idea ;)   All the best! /Marco ... View more

Hello:     I'm looking for an implementation of prediction intervals for the "non-parametric" case, conceptually analogous to the normal parametric case as currently implemented in the Distribution Platform in JMP, as illustrated by the following demo by @julian per the JMP STIPS Course. https://community.jmp.com/t5/Short-Videos/Calculating-Prediction-and-Tolerance-Intervals/tac-p/409311#M236   I have a good textbook reference for this but the method is almost strictly table-based, with some Fortran code for the nonparametric intervals:  Gerald Hahn and William Meeker, Statistical Intervals: A Guide for Practitioners 1991, Wiley & Sons   Why ask for such a "nuanced" thing??      In many situations we are required to use non-parametric methods or otherwise transform the data to normal if the sample data indicates non-normality on the basis of the Shapiro-Wilk Goodness-of-Fit test. This is a procedural requirement for some businesses despite the pros and cons of proceduralizing (since in many cases the normal approximation will work just fine if one looks at the assumptions and the conceptual framework of the interval estimation method or the hypothesis testing method).   Thank you, @PatrickGiuliano  ... View more

Hello, I have some feedback for FDE usage:   I would like to set the default values for the set of grids and spline degree for the splines (see picture). In the preferences I do not an option to do this.   Being able to control this over preferences or the the model control itself (like a apply to all button) would be an ease for using FDE.     Thanks! Ben ... View more

I have one request/recommendation for a future release of JMP.  In the one-sample equivalence test platform in JMP 16.0.0, I noticed that there is no correction for the non-parametric test case.  Yes I can, in essence conduct the test using the one-sample T-test platform: and by running the Test mean option 2X and picking off the appropriate tests as shown in the graphic below:   But I can’t get the Equivalence region nor can I get the TOST framework formalized in this same way which is what I like about the Test Equivalence Platform for the normality assumed case. So formally the request is for a non-parametric version of the TOST for the one-sample test.    Note that for the 2-sample case I can compute a non-parametric confidence interval on the median difference and compare that to the practical difference region and effectively run the TOST in this way (e.g. Hodges-Lehmann confidence interval on median difference using Wilcoxon all-pairs NP test). This is just the “confidence interval version” of the TOST.  So this actually works quite well for the 2-sample case without any need for an additional feature in the Equivalence Test platform in Fit Y by X.   But in the 1-sample case, in the Distribution platform, I cannot do the same, why? The only way I know of doing something similar to this would be using Custom Quantiles in  JMP's Distribution platform, and that does provide a non-parametric estimate (albeit not the Hodges-Lehmann one, and I have trouble specifying the exact desired coverage for Confidence level of the TOST which is a standard 90% for a 5% significance level).   So (lower-priority) an additional feature altogether could be, the calculation of a non-parametric confidence interval estimate using Hodges-Lehmann on the median in the Distribution Platform. Or perhaps a boot-strapped confidence interval (which you can do for almost every table already in JMP Pro with Right Click > Bootstrap on data table output).   ... View more

Let's say you are running a One-way ANOVA with 5 groups.  You test for equal variance and conclude that the assumption of equal variance does not hold.  You then evaluate Welch's test.  The p-value from Welch's test indicates that the 5 group means are not equal.  Is there a simple way to run a multiple comparison test (to compare the 5 means or a Dunnett's like test to compare 4 of the means to a control group) when you have unequal variance?    Ideally, this would be an option found in the "red triangle" when using Fit Y by X (for the One-way ANOVA case).  Perhaps found in the Compare Means choices (see attached for a possible example).   Thanks ... View more

Hello:   Can future releases of JMP include a confidence interval for reliability estimates (like McDonald’s Omega) that is being reported in the SEM platform?  This will be useful to know besides just a point estimate. ... View more

I wish there was a bayesian analysis platform in JMP. Also, if there was an option to do a Bayesian analysis from an existing result. i.e. start with a chi square p val, and then use that to create a bayesian probability analysis where we can choose ranges of priors, and see what the posteriors would be. More and more Bayesian analysis is being recognized as a powerful, informative, and many times more "correct" way of understanding than the usual hypothesis testing and p values. I feel the Bayesian world is largely missing from JMP. Beyong adding a specific platform, the ability to do Bayesian analysis of an existing result (like simulation or bootstrapping are now) would be extremely useful and powerful. ... View more

I'd like to have the ability for the minimum split size for tree models to be based upon a % of the training sample size vs a integer number.  So for example I could ask that the minimum split size would be 10% of the training samples size, rounded up. ... View more

I'd like to have a multivariate x axis in the Fit Y by X platform, with the ability to run Means testing. I use the Variability chart in the variability gauge platform to show how multiple (non-continuous) variables grouped on the X axis impact a Y numeric product response. I'd like to be able to run means ANOVA testing for each of the variables plotted on the X-Axis.  ... View more

It would be helpful to able to generate a correlation matrix using nonparametric methods through the Multivariate Analysis tools.  The current matrix that appears by default uses the Pearson product-moment method.  The display for nonparametric statistics is completely different, in a long pair-wise list of all combinations.  (see graphic below)  It would be nice to have a nicely formatted matrix for different correlation types.  We shouldn't use Pearson correlation for everything, right?  ;)     It would be especially nice if other options, like correlation probability matrix, was also available for nonparametric methods.  Perhaps the default correlation method could be listed on the main Multivariate interface (see graphic below).       ... View more

I know that, a while back, Jeff P, explained a longer work-around for this, but I think it would be great to have a GraphBuilder option in which Tukey's letters (to denote differences between plotted means) could be overlaid over columns/icons in the plot itself. Normally, I just paste the JMP plot into Adobe Illustrator and manually add them (see attached image), but this can be pretty time consuming for a complex figure. I could envision it as an option on the left side (near where you define error terms and things of that nature). I guess you would need to somehow define the treatment you are interested in, which may be incompatible with certain graphics, so I can kind of see why this may not currently exist (i.e., there are soooo many plots that could be generated, only a fraction of which would make sense to use with Tukey's letters over the data points).  ... View more

I have periodic inprocess record for batches manufactured. I need to calculate batch specific mode for the respective in-process record. I can't see the option of mode statistics in Analyze > tabulate. Rather than calculating it from raw data and populate to data table and then in tabular summary using analyze>tabulate option, it would be handy and time effective if the Mode statistics is default available in in tabulate. Thanks!! ... View more