JMP Wish List

What inspired this wish list request?   Models in the Predictive Modeling platform do not have a default capability to produce confidence intervals or prediction intervals around model predictions; however, JPM has provided a way to generate confidence intervals around the predictions in the profiler using bagging (https://www.jmp.com/support/help/en/18.0/#page/jmp/example-of-confidence-intervals-using-bagging.shtml#). Unfortunately, there does not seem to be a way to produce prediction intervals in the profiler for predictive models using JMP (I am a JMP 17 user so I am unsure if this capability has been added, but based on the JMP 18 help entry I don't believe this capability has been added). I suspect it is feasible to produce these prediction intervals based on my understanding of how the confidence intervals are calculated.     What is the improvement you would like to see?    With JMP 18's renewed its focus on prediction intervals in the profiler (https://community.jmp.com/t5/JMPer-Cable/Prediction-Profiler-enhancements-in-JMP-18/ba-p/718342), my hope is that prediction intervals for predictive models will be added. It would also be helpful to generate column formulas for confidence intervals and prediction intervals for predictive models.   Why is this idea important?   Prediction intervals are useful to judge whether the data generated by the proposed model looks like the actual data, often called a "posterior predictive check" in the Bayesian literature. If the data generated by the model does not resemble the actual data, then it may be desirable to consider a different model. Please see a description of this concept, along with a simple example, from John Kruschke's book "Doing Bayesian Data Analysis" below:   "The fifth step is to check that the model, with its most credible parameter values, actually mimics the data reasonably well. This is called a posterior predictive check... One approach is to plot a summary of predicted data from the model against the actual data... The predicted weight values are summarized by vertical bars that show the range of the 95% most credible predicted weight values. The dot at the middle of each bar shows the mean of the predicted weight values. By visual inspection of the graph, we can see that the actual data appear to be well described by the predicted data. The actual data do not appear to deviate systematically from the trend or band predicted from the model. If the actual data did appear to deviate systematically from the predicted form, then we could contemplate alternative descriptive models. For example, the actual data might appear to have a nonlinear trend. In that case, we could expand the model to include nonlinear trends... We could also examine the distributional properties of the data. For example, if the data appear to have outliers relative to what is predicted by a normal distribution, we could change the model to use a heavy-tailed distribution..."         ... View more

What inspired this wish list request?  I have been working on predictive models where I have my data set split in to defined inputs and a validation split, but I have up to 9 response parameters I am trying to model. I noticed if I make a PLSR model I can add all the response parameters to the response box in the gui and the software will make 9 independent models (one for each response parameter) summarized together in the results. But when using an algorithm like a NNET I can either model each response parameter one at a time, or if I had them all to the response box in the gui instead of JMP speeding up the process and training a NNET for each parameter, it trains a single combined NNET model to try and predict each output. It simply is not clear when the software is making independent models or fusion model.   What is the improvement you would like to see?  I think there are two additions that user would find helpful.  Updating instructions in the how tos It would be good to have it clarified in the help/how to section how JMP treats modelling multiple response. Adding a radial option in model gui for the applicable algorithms. I think the easies option would be to include a radial option on the amenable algorithms (NNET and Tree-based etc.) that would let the user select if they want all the response parameters to be modelled collectively vs. independently.       I think there was an option similar to this for how jmp handles multiple response parameters back in JMP 13 (maybe still included in JMP 15) but its an option that seem to have been removed.       Why is this idea important?  I think this idea is important for two reasons.  First is simple clarification. It wasn't immediately evident how some of the algorithms like NNET treat multiple response parameters in JMP and I think users would benefit from a more straightforward description of how modelling multiple responses in JMP works so they know the software is doing what is expected.  Additionally I think this option, just gives users more flexibility in how they use JMP efficiently and would allow them to maximize the value they gain from JMP. In my experience their are applications where modelling results together vs. independently can be critical to maximize model performance.    ... View more

What inspired this wish list request?  Trying to make slight modifications to my existing models when I just have formula columns seems to take quite a bit of effort unless I have the script every time saved.   What is the improvement you would like to see?  In most simple case I would be happy to just have a way of getting all the column names used in a formula on clipboard (also make this scriptable without clipboard...). There might be other wish list item for this item but I could quickly find one. There are solutions in the community for this but they are fairly complicated Head Arg Recurse Formula , How to automatically extract the column name used by the formula in the JSL code? .   In best case I would like to be able to re-launch the platform which did create my prediction columns. As I'm JMP Pro user, I would be fine having this as a feature in Formula Depot. In this case it would be nice if different settings would be also pre-filled IF they are stored to the column properties.   Why is this idea important?  Makes it easier to go from prediction formula back to the launcher (when possible) to make small modifications if needed or just to rerun the model with same settings without having to save the platform script every time.   ... View more

What inspired this wish list request?  Unlike other statistical software (Minitab), it seems that JMP doesn't provide detailed formulas for analyses. This lack of details and transparency can make it difficult for users to understand the underlying calculations.     What is the improvement you would like to see?  Provide more formulas for analyses where it makes sense to include them in the support documentation and in cases where formulas are quite complex, simplified formulas for simplified examples so the user can understand the concept.   Why is this idea important?  understand the underlying calculations, it is quite often to get request to explain the analysis, how do you derive this conclusion and what kind of analysis and fundamental concept formulas are being used in the analysis.   ... View more

Hello, I know there has been similar posts and requests on this topic, sorry about that. Some applications (e.g. in spectroscopy analysis) require doing signal deconvolution using a simple combination of multiple gaussians (see capture attached). I have seen alternatives suggested here (by using the distribution plateform, or with PLS). Does it mean that the multiple gaussians fit method will never be implemeted in JMP (or in a custom add-in)? Thanks Best Regards    ... View more

What inspired this wish list request?  (1) From 'Functional DOE Profiler' feacture in Functional Data Explorer in JMP Pro, I would like to see   (1-1) Confidence Intervals   and   (1-2) Interactive HTML file export from the profiler   (2) From 'Model Comparison' in JMP Pro,    (2-1) 'Model Averaging' (Ensemble) result showed a fixed RSquare (1.00) and RASE (0.00). I think it's a bug.     (2-2) Shelf made 'Ensemble model' that is formulated from other models (Least Squares, Partial Least Squares, SVEM Lasso, XGBoost, Neural Turned with different weight) also showed a fixed RSquare (1.00) and RASE (0.00) value in Model Comparison results.    What is the improvement you would like to see? Why is this idea important?      ... View more

Hello,   In Fit Model, macros (e.g., factorial, response surface, etc.) only work when columns in the Columns list are selected. It would be extremely convenient and helpful to allow the same functionality by selecting columns in the Effects / Factors box.   Thanks and regards Mehul Shroff ... View more

What inspired this wish list request?  I encountered the situation when trying to realize classification with several Machine Learning algorithms on Mixture Space-Filling DoE data (presentation here : https://community.jmp.com/t5/Online-Abstracts/Synergy-Between-Design-of-Experiments-and-Machine-Learning-for/ec-p/708147#M2). In order to present the results in a visual and clear way, I wanted to graph ternary plots with continuous responses expressing the change in classes, so I opted for the favorite class probability. But I had to do a workaround to actually create the graph, as the Ternary plot requires a formula column, so I did it in two steps : Create a classification model (SVM) and click on "Save Prediction formula". Class prediction is a formula column, but not the calculated probabilities columns for each class. Create a model with the same inputs variables / X's, but the response is the favorite class probabilities column. I can then save the probability formula and use it in the Ternary plot.    What is the improvement you would like to see? As in the model platform the calculation of the class probabilities is already done in order to predict the different classes, I would like to have the calculated class probabilities saved as formula columns in the datatable. This would be a more straightforward option, require less workarounds, and would avoid any model misspecifications in the second step.       Why is this idea important? Ternary plot or Surface Plot are essential tools to better visualize and apprehend RSM/Mixture models behaviors. With categorical responses/classes, it's difficult to represent correctly and continuously the variation in the experimental space, and I think using favorite class calculated probabilities formula could help enhance understand the influence and importance of factors on the categorical response(s).     ... View more

What inspired this wish list request? On a couple of webinars showing the Predictor Screening, I have seen the user add some random normal data, to show where just random data might fall within the Predictor Screening output. This seems to be a great idea to always have that option, or even make it the default.    What is the improvement you would like to see? Add the option on the launch dialog to add Random Normal data and/or Random Nominal data to allow comparison to just random chance from continuous or nominal predictors. Might be an option down near the "Number Trees" and "Set Random Seed", or under the "Cast Selected Columns into Roles".  Ideally would color code in the table output to highlight "Random Normal" and "Random Nominal N".  The random nominal may be more difficult. In the JMP sample "Bands Data", some of the predictors are 2 categories, or 3, 4, 5. So either give user the chance to specify, or maybe there is a procedure to add 2,3,4,5 if already exist in the predictors.     Why is this idea important? Ease of use and interpretation of the results. I like how p-value is color coded and sorted in many of the platform outputs, this would be similar to show about where random noise would be in the Predictor Screening results.     ... View more

What inspired this wish list request? I am using FDE to create FPC Scores to use as inputs to into a Generalized Regression model.  I would like to use this model outside of JMP Pro as part of a quality Control Process.  While I can use the GenReg model outside of JMP, I have no way to generate FPC Scores from the time series to use as input.     What is the improvement you would like to see? I would like to be able to to export code that takes the time series as input and exports FPC Scores.  If that is too difficult, at least create a way to generate FPC scores from new data inside of a JMP sheet.  Ideally both solutions.     Why is this idea important? This idea is important because while creating a solution in JMP is great, being able to use the solution in a quality control process is necessary.     ... View more

I was analyzing a bunch of historical data for engineers who had several potential multiple response column factors. They have only regular JMP (not JMP Pro), so several steps were needed to make them a Profiler they could play with in their copy of regular JMP.   JMP17 and JMP18EA8: I discovered that I could not get a profiler to appear in GenReg with the multiple response column factor in the model, much less if I saved the prediction formula (and PredSE formula) and used Graph>Profiler. However, if I edited the prediction formula column to remove ContainsItem() and replaces with If(Contains(),1,0) and replace the multiple response column reference with a nominal column, then Graph>Profiler would work fine. And I could give my engineers a Profiler they could play with. But these are a lot of annoying extra steps. Save me the steps!      ... View more

What inspired this wish list request?  Need to vizualize in a ternary plot big amounts of mixture data generated from Monte Carlo simulation     What is the improvement you would like to see?  Need in ternary graph same density contour plots as already implemented for other types of graphes    Why is this idea important?  This would greatly improve the vizualization of big data for mixture application     ... View more

Waveform analysis is being performed on parallel data lanes.  For each data lane when the Functional DOE Analysis option is selected two curves are produced.  The first, a mean curve and the second is a sloped line.  This is great!  I believe a nice capability would be another summary option that plots these same curves for each data lane alongside (or above/below) each other.  I have found "identical" data lanes are not identical and easily viewing the data lane to data lane difference is helpful while trying to resolve test yield issues.  My work around is to copy and paste the thumbnail plots into a separate document.   Here's an example of just one data lane.  I would like to view multiple plots like this for each data lane side by side.     Using JMP Pro 17.2.0     ... View more

What inspired this wish list request?  I wanted to compute the uncertainty of prediction fastly and easily for predictive model without using bagging. Indeed, bagging is quite slow to compute with the profiler.   What is the improvement you would like to see?    I would like to save the all the trees formula when I use the Bootstrap Forest Plateform. Or at least the standard deviation of the distribution.     Why is this idea important?    This is very important for me to be able to compute acquisition function when I do Surrogate Model Optimization (Bayesian Optimization). I tried to use bagging with the partition plateform, but the result is not satisfactory. The gaussien process plateform do not deal with categorical variable.     ... View more

What inspired this wish list request? JMP is a great tool for data exploration, and helps build quickly prototypes for predictive models, to see if there is an added value in pushing the analysis and modeling further. Looking at the panel "Overall Statistics" for a classification topic, I find there may be some performance metrics missing that are frequently used in the Machine Learning / Data Science world. Here are the metrics currently displayed for a classification topic (example from a bootstrap forest):    What is the improvement you would like to see? I would like to see displayed commonly used performance metrics for classification, such as Sensivity, Specificity, Precision, Accuracy, F1 score, Matthews Correlation Coefficient ... Perhaps directly under the Confusion matrix ? Here are the possible metrics in use for classification topics from an online calculator, with the formulas explained : Confusion Matrix - Online Calculator (onlineconfusionmatrix.com)   Why is this idea important? As JMP is trying to bridge the gap between the Statistics and Machine Learning worlds by integrating several models from these two modeling mindsets in one platform, it would make sense to also integrate Machine Learning performance metrics in the results of a classification model, in order to have all infos in the same place and be able to speak the same common language.  ... View more

Hello,   Can the antecedent, consequent, and rule sizes be included in the output table from the Association Analysis platform? Thanks Mehul Shroff ... View more

What inspired this wish list request?  GAMs have become so popular, but JMP doesn't have them.   What is the improvement you would like to see?  I would love to see JMP have GAM capability in their predictive platform   Why is this idea important?  GAMs are becoming the "go to" method for a lot of analysts and JMP would be a much better product IMO if they created this functionality. I see requests going back to 2018. GAMs are more and more popular.   ... View more

What inspired this wish list request?  When building complex models based on large number of collinear/correlated variables, there are several possibilities available. One of them is transforming the variables into principal components, and then create a model based on these principal components to avoid collinearity issues, reduce dimensionality and keep a large part of the initial information. The problem that could happen in predictive modeling is data leakage : information from the validation set and/or test set is used in the pre-processing and/or model building. This can happen in PCA platform if validation/test rows are not hidden and excluded : if the PCA is done on the whole dataset and after models are built on the training principal components data, the model still contains a part of the information from the validation and test sets. Model evaluation can become biased as the performances would be assessed too "optimistically".   What is the improvement you would like to see?  As for other platforms (see Fit Model below), I would like a validation role when launching the PCA platform dialog, so that the PCA is created with the training rows, and applied on validation and test rows :      Why is this idea important?  The validation role for PCA would enable user to be more aware of this data leakage situation when they are building predictive models. It's also part of the best practices to split the dataset first into train/valid/test sets before doing any transformation/model, so adding this validation role would enforce a reliable best practice for users. Finally, it would be also easier to use PCA in a more reliable way as a pre-processing step for predictive models with this feature, instead of manually or scripting excluding rows.   ... View more

What inspired this wish list request?  StepWise Fit incorporates statistically invalid correlations into the model fit when there are a large number of parameters which are fit. JMP Stepwise criteria is only looking at Sum of Squares along with the F Ratio to produce what stays in the model.  Then only when you Make and Run the final Stepwise Model do you get the Fit Group Effect Summary that gives you the option to get FDR Logworth and FDR PValues. An FDR alpha should be the default input for StepWise rather than the current inputs.    We had to research why JMP is failing to select the correct criteria for incorporating parameters into the fit. Project Euclid (https://projecteuclid.orgtalking)  talks about incorporating FDR in Stepwise and we implemented into Python since JMP could not perform this properly.      What is the improvement you would like to see?      Why is this idea important?      ... View more