Change x axis of a hotelling T2 from sample to a date time sequence

[#1]

How do I replace the x-axis of a multivariate control chart? Default it shows sample number

[#2]

Hi @MarcP ,

  I'm not sure if that is possible in the Model Driven Multivariate Control Chart platform, which it sounds like you are using.

  You can however do it an alternate way. If you take your columns of interest and run them through the Principal Component platform, you can perform an Outlier Analysis (red hot button > Outlier Analysis). You should check the minimum number of components you need by PC red hot button > Eigen Values; rule of thumb is the minimum number of PCs you need is what can explain >= 85% of the data. Then, for the Outlier Analysis red hot button > Number of Components, you can then enter in the number of PCs needed. Again, go to Outlier Analysis red hot button and select Save T^2, and it will save a formula to your data table. You can then do a Graph Builder with your T^2 on the Y-axis and your date/time column on the X.

  I just checked, and under the MDMCC platform, you can go to the T^2 red hot button and save columns, and it will save the T^2 formula as a new column in your data table. You can then do a Graph Builder plot as I mentioned above.

Hope this helps!,

DS

[#3]

Hi @SDF1 ,

Thanks for your answer. How would you go about if I wanted to use subset A to calculate the T2 targets and use your suggestion on subset B and add/embed the target results in the graph?

[#4]

Hi @MarcP ,

  I'm not sure I understand your question entirely. Could you show an example from your data or from a sample data table, like the Semiconductor Capability table in the JMP sample data directory?

Thanks!,

DS

[#5 · Accepted solution]

Just to verify - you are using the Model Drive Multivariate Control Chart in JMP 16?  Have you tried the Historical Data End at Row option?  If you have subset A as the first x rows in the table, then set Historical Data End at Row to the x rows of subset A, then it will calculate separate limits based on subset A versus subset B. 

[#6]

@ledi_trutna , no I actually was using the Analyze --> Multivariate methods --> Principal component analysis option.

Yes, I do have JMP16