nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Best statistical method for analyzing data set with a manufacturing change

VariancePony864

Hello- I am working with the following data set. Column1 (#s 1-47) are lot numbers; Columns 2,3,4 are scientific results from an analysis. The rows highlighted represent a change that took place in our manufacturing process in which a new lot of a raw material was implemented. My goal is to understand if a change in the raw material lot has an impact on the analytical results. What statistical method(s) would you recommend to best deomnastrate this?

JMP file.xlsx

Accepted answers

statman

I opened both your excel file (to see what rows you highlighted) and your JMP file. They are not identical, so I used your excel file to create the appropriate JMP file. I marked the rows you highlighted with an X. If you open the file and click on the green arrows in the left hand zone of the file (labeled IR by lot and Multivariate). These are 2 analysis of the data set. I don't see any evidence the lot changes are unusual, although there are some other interesting points in the data set.

JMP file1.jmp

Victor_G

Hi @VariancePony864,

And welcome to the Community !
To add one graphical analysis suggestion based on the excellent comment by @statman, it could also be possible to use Model Driven Multivariate Control Charts (jmp.com), in order to have one chart taking into account your 4 (correlated) responses from your process.
This way, you could proceed with your analysis/inspection in two steps :

Take a look at the Multivariate Control Chart to see if there is any unusual pattern or points (in a synthetic and global/macro view),
For these potential unusual data points, see what are the responses most contributing to the deviation (just put your mouse over a datapoint to see the repartition of this "deviation" by responses), and have a look at individual response with the control charts recommended by statman.

Attached you'll find the datatable created by statman with a new script added "PCA Model Driven Multivariate Control Chart".
Hope this answer will help you,

JMP file1.jmp

All comments

statman

Welcome to the community. There are multiple ways today this (not sure what "best" means), but I prefer graphical using control charts. It looks like you attached an Excel file. Could you attach a JMP file and we can help format it correctly?

VariancePony864

sure thing, see attached, thank you!

Take a look at the Multivariate Control Chart to see if there is any unusual pattern or points (in a synthetic and global/macro view),
For these potential unusual data points, see what are the responses most contributing to the deviation (just put your mouse over a datapoint to see the repartition of this "deviation" by responses), and have a look at individual response with the control charts recommended by statman.

Attached you'll find the datatable created by statman with a new script added "PCA Model Driven Multivariate Control Chart".
Hope this answer will help you,

JMP file1.jmp

VariancePony864

Thank you both so much for your help! It is much apprecaited!

stefensmith614

Great question! When you have a manufacturing process change—like introducing a new raw material lot—and you want to see if it impacts your analytical results, the approach usually depends on whether your data is normally distributed and the sample size.

Here are some commonly used statistical methods:

Two-Sample t-Test (or Welch’s t-Test)
If your data is approximately normal and you have two groups (before and after the change), this test compares the means of the two groups.
Mann-Whitney U Test
If your data isn’t normally distributed, this non-parametric test is a good alternative to the t-test.
ANOVA (Analysis of Variance)
If you have more than two groups or want to analyze multiple factors, ANOVA can help identify if there’s a significant difference across groups.
Regression Analysis
If you want to control for other variables or see the magnitude of the change, regression is powerful—especially when combined with interaction terms.

In the Industrial IoT and digital manufacturing space, companies like Siemens and INS3 often use these statistical techniques embedded in advanced analytics platforms. These tools not only run the tests but also integrate real-time process data, which helps detect subtle shifts in quality early on.

If you have historical data, you could even implement control charts or CUSUM for ongoing monitoring—great for continuous process improvement.