JMP Wish List

This is probably the most frequent limitation of JMP at my organization.  Unequal variances suggest to use Welch's ANOVA, but options are limited for multiple comparisons with unequal variances.  Steel-Dwass doesn't work well with small sample sizes and it's not clear that it is robust to very unequal variances. Games-Howell test is effectively a Welch's version of Tukey's HSD and works well with small sample sizes. ... View more

I find on larger data sets that some analyses take a long time to run.  For example, running a Binomial or Poisson generalized linear model on a data set with ~20 million rows can take over an hour on my machine.  It would be great if JMP could (1) warm me if executing a command was going to take longer than 10 minutes and give me the option to cancel; and (2) if I choose to continue, give me an estimate of how long the analysis was going to take to complete. ... 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

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

Highly regulated fields such as medical devices or pharmaceutical require that a statistical rationale for any design verification and validation activity is used. If a variable is continuous, the sample size for tolerance interval is a calculation based on the type of interval (1 vs. 2 tails), confidence, population coverage and margin of error. If a variable is an attribute, typically a count of conforming items, the calculation's output is an acceptance number (maximum number of non-conforming items that can be accepted in the experiment) and a sample size, based on confidence and population coverage.   Sometimes even the conformity of a product is defined on a such statistical basis, sometimes with numbers which are impossible to find in tables: Les contrôles des leucocytes résiduels dans les PSL sont effectués dans un délai maximal de 24 heures après la déleucocytation. De plus, les contrôles de chaque type de PSL, dans chaque établissement de transfusion sanguine (ETS), doivent montrer que, pour un calcul fait avec un degré de confiance de 95 % et en utilisant un plan d'échantillonnage adapté, les contenus ci-dessus doivent être respectés au minimum dans : 97 % de la production des PSL cellulaires déleucocytés ; 95 % de la production des plasmas à usage thérapeutique déleucocytés. (e.g. allowed white blood cells post filtration in France, 97% conformity with 95% confidence see link https://www.legifrance.gouv.fr/affichTexte.do?cidTexte=JORFTEXT000023140529 ) So these tools are common also in PCMF (post market clinical follow-up) activities or market monitoring.   It would be nice to see this topic covered in JMP.   Some links: http://reliabilityanalyticstoolkit.appspot.com/sample_size https://support.minitab.com/en-us/minitab/18/help-and-how-to/statistics/power-and-sample-size/how-to/sample-size/sample-size-for-tolerance-intervals/before-you-start/example/ https://www.qualitymag.com/articles/91991-sample-sizes-how-many-do-i-need http://www.cbinet.com/sites/default/files/files/Session12_Golden_Alan_pres.pdf https://www.mddionline.com/sample-size-selection-using-margin-error-approach   ... View more

☐ cool new feature ☑ could help all Jmp users! ☑ removes something that feels like a „bug“ ☐ nice to have ☐ nobody needs it   What inspired this wish list request?  In Jmp18, a warning message appears for EVERY user - EVERY day - EVERY time Jmp is started  ... unfortunately: already months before Jmp expires. Unlike a splash window, the pop-up window remains in place until the user closes it manually.   This is how it looks like:   What is the improvement you would like to see?  Please - remove the warning message [either: completely, or: after 3 seconds] - just show the warning message 1 - 2 x before it gets disabled [automatically] or: - add an option to disable the warning message (like for the startup splash window).   Why is this idea important?  less confusion for JMP users no annoyed license site managers   other wishes from    TS-00143027   ... View more

In manufacturing is very common to handle datasets with more than a hundred columns, each corresponding to a sensor. Manufacturing systems have a tree hierarchy that is defined in PI AF (also used in JMP 17 and PI to find the sensor names).   This information is lost when moved to a JMP table.  We open-sourced an add-in to reorganize columns (see image and code here)   However, JMP only allows two levels (parent group and child), which is very limiting in terms of organization and analysis. For manufacturing data, multiple grouping levels are a must-have or users won't be able to use JMP for their daily tasks in manufacturing (too much information to handle).   Please have a look at PI AF to understand how engineers are used to finding data. This information can be coded into a column property. This metadata can eventually allow users to swap analysis between assets. ... View more

Isolation Forest has been emerging as arguably one of the most popular anomaly detectors in recent years due to its general effectiveness across different benchmarks and strong scalability. It is computationally efficient and has been proven to be very effective in anomaly detection.   Isolation forest compared to others (src) The algorithm isolates observations by randomly selecting a feature and then randomly selecting a split value between the maximum and minimum values of the selected feature. The implementation is therefore the same as partition trees and bootstrap trees with the difference that the target function should be uniform (trying to split variables with a constant Y value). Yet, if you try this in JMP, it doesn’t work. By including Isolation Forest in JMP, similar tu predictor screening, users would have access to a powerful tool for detecting anomalies in their data. This could help them identify unusual patterns or behaviors that may warrant further investigation.   This paper studies how IForest works and improves upon its few limitations (i.e., extended isolation forest) https://hal.science/hal-03537102/document   Scikit-learn documentation https://scikit-learn.org/stable/modules/outlier_detection.html     ... View more

I will occasionally have a distribution with zero's in there and when I try using the Distribution platform, I get a much smaller list of possible distributions to fit than normally.  I then can go into Life Distribution and look at a ZI Lognormal or ZI Weibull, but why not offer these in the Distribution platform? ... View more

What inspired this wish list request?  Few posts I found from community and pinning ruler could also be fairly helpful from time to time.   What is the improvement you would like to see?  Ability to pin rulers and crosshairs to graphs (with crosshairs it should be possible to hide the aiming circle). Could also be helpful to be able to add multiple of them and possibly have some additional settings such as colors (for ruler also changing increment).   Why is this idea important?  Provides additional tools to annotate your graphs   ... View more

We teach many of our engineers about paired t-tests and use the Fit Y by X platform (rather than Matched Pairs) when the format of their data set is more suitable for Fit Y by X. There are 2 frustrating aspects:   There is still an option for 't test' in the analysis, which does an ordinary two-sample t-test, which ignores the Block that you incorporated. Most engineers/scientists don't realize this. If the data is unbalanced by even 1 data point (e.g. I have 20 pairs, but there is 1 missing data point or outlier that was discarded), you get an error message saying the Blocking is going to be ignored. It's a somewhat confusing message for them, and I feel that most engineers/scientists just blindly keep following the platform output. If I had fit the same model using Fit Model, this slight unbalance in the data can easily be utilized to give the correct hypothesis testing output. I'd like to see the Fit Y by X output handle the unbalanced data when the correct model is rather straightforward. ... View more

I'm a little underwhelmed at times with the options available under the DOE --> Design Diagnostics --> Sample Size and Power menu. Two particular omissions seem noteworthy for what I do: 2 standard deviations Equivalence tests (one, two, paired samples) In some other threads I've seen some mention around the topics, but they are not in the wish list.   Here are some links to those threads: Sample Size and Power for Testing standard deviations  TOST Acceptance criteria and Sample Size  As a reference, Minitab does seem to provide these options. I've use proc power in SAS for these situations. ... View more

Our company is now using automated scripts in JMP connecting to our Laboratory management system, LIMS, to pull data for SQC reports.  These need to be reviewed for 'bad data' prior to sending to customers.  Samples undergo multiple tests which need to be arranged in columns - 1 row per sample.  Most of the 'bad data' relates to a single test (maybe a typo or a measurement special cause) so we don't want to exclude the whole suite of testing just to remove that single point. JMP's row structure to data currently means the only way this is possible is to exclude values in column properties or delete the value from the cell.   ... View more

I would like an option in the Distribution Platform Preferences to set standard values for Confidence and Proportion to Cover to be used in Tolerance Interval Calculation.   Similarly, when computing Tolerance Intervals for many by-groups, it would be great if there was an option to specify Confidence and Proportion to Cover only once for all by-groups, because typically in such a situation I want to use the same levels for all groups (or is there a way to this by holding down a key or sth? Using ctrl works only to the degree that the interval is computed for all groups, but each time levels must be specified). Maybe you can supply typical settings as you do for Confidence Intervals, although here it would have to be 3x3 (0.9/0.9, 0.9/0.95, 0.9/0.99, 0.95/0.9, .... ) plus other...   ... View more

Please add option for mediation analysis in JMP. It can be done in SAS using PROC CAUSALMED. ... View more

Row functions are essential in manufacturing data.  We use Dif() to calculate the rate of change (slopes, or derivatives), which can be correlated to our target of interest. Lag() is also used to align data (upstream and downstream equipment)   However, when having tables with a column indicating different groups, new users are out of luck.   Two consecutive rows can belong to different groups and the only way to guarantee this is by adding complex formulas with conditionals (example).   It will be much easier for the user to include a <byVar> argument in Dif() and Lag().   For example, Col Cumulative Sum() is another row function that does allow this!   Here, I am creating a simple counter where I don't need to be worried about conditionals.   ... View more

☐ cool new feature ☑ could help many users! ☑ removes something that feels like a „bug“ ☐ nice to have ☐ nobody needs it   Short: Graph builder needs this function:       Long: What inspired this wish list request?  Every Jmp user knows this issue: when you invested 15 minutes to optimize your Graph Builder plot, added some subplots, added some categories and then the legend looks like this - and you need another 15 minutes to figure out which entry in the legend corresponds to which subplot.   When computers started to have several video outputs, computer users had a similar issue: which monitor setting belongs to which monitor ?!?!?   And the programmers had a great invention: an Identify button. The user just has to click the button and numbers will appear: 1) in the overview 2) on each monitor   What is the improvement you would like to see?  1) I want the same functionality for Graph builder. An Identify Button . When a user clicks the button, numbers should appear in the Frameboxes and in the legend: [extreme case to illustrate the unbelievalble help of the suggested function]   ") please also add a button to "reset" the sequence of entries in the legend to a default setting: all entries from the first subplot, then all entries from the second subplot ....   Why is this idea important?  Nowadays it's very difficult to use different color coding in different subplots - as soon as a new subplot is added, all the color settings are scrambled. The unpredictable sequence of entries in the legend makes it almost impossible to efficiently adjust the settings.         more wishes from  ... View more

When performing certain tests in JMP, the null hypothesis is stated, but not for all.  I found several people are confused by the interpretation of the results for hypothesis tests if the null hypothesis is not stated.  Could a future version of JMP state the null hypothesis whenever there is one to consider? ... 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