About tnad

tnad · ‎03-04-2020

Thanks. I can filter on p-value here, but I'm not sure how I can only keep variables that are not highly correlated with each other using this table

tnad · ‎03-04-2020

I'm screening through thousands of variables and would like to only keep ones that are not very highly correlated with each other for example: r2<0.95 and >-0.95. Is there an easy way to do this? I can use: "Multivariate methods > multivariate" to calculate r2, but I have no idea how I can make the selection according to cutoff above.

tnad · ‎03-02-2020

I have a small data set with 154 molecules (attached). I'm trying to predict KV40 using 6 factors. From looking at similar studies in the literature, many use the leave-one-out validation method to build models for such smaller data sets, so to replicate this in JMP, I did the following: Fit-model: Stepwise regression - used response surface for factors - used k (k-fold cross-validation) = number of samples = 154 - left everything else as default Result: r2 = 0.84; r2 (k-fold) = 0.72 After box-cox transformation: r2 = 0.87 (r2_adj = 0.86) I tried to do this with a validation column (70:30) and did not get a result as good as this. I have JMP pro. My concern is am I creating a fake or overfit model when I do this type of cross-validation? Is there a better way to do this? Is there anything I should watch or test for? Thanks.

tnad · ‎10-18-2018

Thank you Mark. This helps a lot.

tnad · ‎10-18-2018

oops here it is:

tnad · ‎10-18-2018

Thanks Mark. Note: the duplicates are in the same location (well number) in each plate. Just in case this makes a difference for the analysis. I tried to replicate what you did with the fit model and got the below: I guess this says there's a significant difference between the duplicates in the two plates? are we able to say how much in comparison to well-to-well variation in each plate? Thanks

tnad · ‎10-17-2018

We did measurements in duplicate 96-well plates (so 96 different samples equally split into two plate). We want to estimate variability between the plates. It seems to me that there are two important sources of variability, one related to time since one plate was measured at a different time and so all samples probably degraded equally in that plate and another related to other differences between the duplicates. We just took the % CV for each duplicate and looked at average or median %CV. Is there a better way to look at how the plates are different with JMP? Also, It doesn't seem to me that "1-r2" (uneplained variation) of the correlation between the plates is the right way to calculate this.

tnad · ‎02-28-2018

yes but what makes it different from the test set. It must be used somehow during model building as a check.

tnad · ‎02-27-2018

I know the validation set can be used to inform the model building process and the test set is not touched. For example, in other models, the validation R2 can be used as a stoppage rule. How is the validation set used in neural networks? I can't find any information about this in JMP documentation or online. Thanks.

tnad · ‎11-01-2017

I'm trying to figure out the best way I can share a model of process data with operators at a plant who do not have JMP. I know the easiest way is to use the profiler and save an interactive html that they can use to manipulate the factors and check the response. The problem is standard errors on predictions disappear when I convert to html. Is there a prefered way to do this? maybe in excel? I've been using JMP pro for over a year at my company and we always hit this wall when we try to share and apply our analysis to real operations.

tnad · ‎07-13-2017

Thank you! I figured out from this script that I can simply add the cost column together with the prediction formula column when I launch the profiler.

tnad · ‎07-12-2017

Hi, I built a model, let's say, with two factors X1 and X2 and a response Y. Using a contour profiler, I can see the values of X1 and X2 that result in my target Y on a grid line. In reality, there is a dollar cost associated with each of X1 and X2 and I would like to display that cost for the X1 / X2 values that match my target Y (on the contour grid line). Is there a trick to do this in JMP? I don't see a way to extract data on a contour profiler grid line. I'm not sure if there's another way to associate the dollar costs of factors to the model?

tnad · ‎07-06-2017

Hi, I'm curious about your thoughts on how you pick a model for a custom DOE (two continuos factors for example). I know if my goal is optimization, I choose an RS model and JMP will add runs closer to center of the design space, but when I examine the design points, I always get the feeling that there are many large gaps between the factors levels chosen by JMP (I check using scatterplot). This worries me that I may miss the sweet spot for optimization. If I increase the number of runs, I end up with more replicates without closing these gaps. I found that I will have to try using higher order models (Powers: 4th or 5th) to fill these spaces. I'm a little new to DOE, so I'm trying to figure out if this is the right way to think about this or to just trust that I will not miss the sweet spot with an RSM if I make sure certain metrics (variance, power) in the design evaluation look good. Thanks!

tnad · ‎06-29-2017

Thank you for your help and suggestions. I'll provide you with some extra information on the factors: continuous X1 (0-500) continuous X2 (0-5000) The number of runs we can do are 25 to 50. The goal is to find the best factor settings/regions to minimize a response Y. We also would like some replicates to asses variability better. The reason 10 to 100 is important for X1 is because of historical data showing that this range affects the response the most. A custom design in JMP for an RSM model will not include this range. Treating the factor as discrete numerical seems like a good option. Will treating it as continous during model fitting be fine in that case? I'm not sure if this is a DOE any longer. I hope you can enlighten me. Thanks!

tnad · ‎06-27-2017

Hi, I'm trying to design a DOE for experimenters at my work and I was asked if I can get JMP to include more runs in a specific range of a continuos factor. This factor's values range from 0 to 500 and they think that the range 10 to 100 is most interesting and would like JMP to add more runs within that range than outside it. Is there a way to do this? Any suggestions would help. Thank you.

tnad · ‎06-16-2017

Thanks. That looks great. Is there a way to get these with the standard error or standard deviation into a profiler? Or is the only way to make future predictions of response is by entering factor values in a new row in the table and checking the result? I'm asking because I'll have to share the results within the company and am trying to figure out the best way to do this. I know the profiler can be saved as html.

tnad · ‎06-15-2017

Hi, I'm using JMP pro to create models of some process data that include 11 factors and a response. What I would like to do is predict the value of the response with a confidence interval (or similar statistic) based on specific values of the factors. This is possible to do with the prediction profiler using linear models under the fit model platform. The problem is that the neural network model seems to do much better than others with this data and when I fix the values of the factors, I get a predicted response value with no statistic on how accurate this prediction is (like the confidence intervals in the fit model platform). Is there a way to get more information on this prediction? Thanks.

tnad · ‎05-15-2017

Hi, I have a couple of questions about augmenting DOE designs: 1- How can I design more than one block for the extra runs resulting from the Augment Design platform? JMP seems to only give the option to place them in one block. I can only run a specific number on one day or I might have hard-to-change factors. 2- I'm going to show below an example of my planned DOE for the initial screening run and how I plan to augment it for optimization (RSM). Please let me know if there's a better way to do this or if this looks fine. Note that my final design will proabably be more complicated (includes mixture), so a DSD will not work. Acid type (Categorical; L1, L2, L3, L4, L5, L6) Solvent type ( Categorical; L1, L2, L3, L4) Acid concetration (continuous; -1,1) Solvent concetration (continuous; -1,1) Substrate concetration (continuous; -1,1) After the initial screening (main effects only) run and analysis, I will probably throw out several levels of the acid type and solvent type. I tested this in JMP and it seems that the best way to do this is to delete the rows in the table that correspond to these levels, followed by running the Augment Design platform. Next, I will choose RSM design to get my extra runs. Does this seem reasonable? Thanks.

tnad · ‎06-12-2016

Thank you all for the valuable information. I have some followup questions regarding transforming the data: 1. From what I understand, you transform some columns but not others? Shouldn't all data be transformed the same way for downstream analysis? 2. I have some columns with many zeroes. Log transformations seem to normalize some but convert zeroes into missing data. Not sure if there's an easy way to do this or some guides on how to handle this in JMP. 3. If normalizing the data is not possible or not essential, is there a preferred method or ones to avoid (PCA?) to further analyze such data? I'm guessing the step following the transformations and screening would be using the strongest contributors in the "fit model" platform and comparing models?

tnad · ‎06-10-2016

I'm using JMP pro and working with a large data set of plant process variables that I want to explore how they might affect production or some other performance parameters in the plant. Are there any guides or resources on how I can start with this? I've been going through JMP videos on youtube but they don't go very deep into analysis. S o far, I've only figured out that I should "screen predictors" to reduce the number of factors I'm looking at. Not sure if next would be fitting to regression models or further exploring data with PCA, pairwise correlations...etc or if I need be careful if my data is not normal data or has a lot of zeroes before choosing the right analysis. Also, I saw some nice plots using canonical correspondence analysis (CCA) to explore correlations with data like mine in R. Is there anything similar/better in JMP? I'm new to this kind of analysis, so any ideas you have on how to get started will help. Thanks.

Date Registered	‎06-10-2016 04:43 PM
Date Last Visited	‎03-04-2020 03:14 PM
Total Messages Posted	20
Total Kudos Received	2

Online Status	Offline
Date Last Visited	‎03-04-2020 03:14 PM

Re: variable selection with a correlation cutoff

variable selection with a correlation cutoff

How good is K-fold cross validation for small datasets?

Re: How to look at variability in this data

Re: How to look at variability in this data

Re: How to look at variability in this data

How to look at variability in this data

Re: Neural networks: validation vs test set

Neural networks: validation vs test set

Sharing model profiler results within company

Sharing model profiler results within company

Help on a continuous factor values/range for a DOE

Re: variable selection with a correlation cutoff

variable selection with a correlation cutoff

How good is K-fold cross validation for small datasets?

Re: How to look at variability in this data

Re: How to look at variability in this data

Re: How to look at variability in this data

How to look at variability in this data

Re: Neural networks: validation vs test set

Neural networks: validation vs test set

Sharing model profiler results within company

Re: Adding a dollar cost to model factors

Adding a dollar cost to model factors

Choosing a model for a custom design

Re: Help on a continuous factor values/range for a DOE

Help on a continuous factor values/range for a DOE

Re: model predictions question

model predictions question

Questions about augmenting a DOE design

Re: Exploring correlations in a large dataset

Exploring correlations in a large dataset