BookmarkSubscribeRSS Feed
Benson

Occasional Contributor

Joined:

Jun 4, 2018

How do I perform correlation in JMP but data is Not one-to-one correspondence?

In this case, I  would like to know if there is any correlation between Ra(Copper roughness) and Chemical(HCL). Therefore, I use Graph Builder to draw 2 smoother line as below. I am wondering is there any way to do correlation-coefficint calculation/analysis, and how do I arrange these data?

Graph Builder.png

 

 

1 ACCEPTED SOLUTION

Accepted Solutions
gzmorgan0

Super User

Joined:

Jul 25, 2016

Solution

Re: How do I perform correlation in JMP but data is Not one-to-one correspondence?

I looked at your attached file. There is a 3rd column named SN (tin?).  SN is only 1 when HCL is collected. Ra is collected with SN of 1, 2, 3, 4 or 5.  There was a little variation in RA for different levels of SN, so I subsetted just the rows where SN==1. The data is not collected at the same time of day.  But  I saw no variation by time of day. 

 

JMP correlation needs matched data or imputed data.  Instead of correlation, look at the time trends of Ra and HCL concentration with an exponentially weighted moving average. Ra looks like it has been trending down, while HCL concentation has different periods. Create a column for the day and summarize Ra and HCL by day. Then use Bivariate as Jim suggested. The results are below.  Note the means of the control charts and teh means of the correlation are not identical, since the means on the control charts are using all the data and teh correlation is the mean of the daily means.

 

image.pngimage.png

image.png

3 REPLIES
txnelson

Super User

Joined:

Jun 22, 2012

Re: How do I perform correlation in JMP but data is Not one-to-one correspondence?

You need to use the Fit Y by X Platform

      Analyze==>Fit Y by X

There you will be able to specify a multitude of lines, including the Kernal Smoother.  It will give you the R-Square

smoother.PNG

Jim
gzmorgan0

Super User

Joined:

Jul 25, 2016

Solution

Re: How do I perform correlation in JMP but data is Not one-to-one correspondence?

I looked at your attached file. There is a 3rd column named SN (tin?).  SN is only 1 when HCL is collected. Ra is collected with SN of 1, 2, 3, 4 or 5.  There was a little variation in RA for different levels of SN, so I subsetted just the rows where SN==1. The data is not collected at the same time of day.  But  I saw no variation by time of day. 

 

JMP correlation needs matched data or imputed data.  Instead of correlation, look at the time trends of Ra and HCL concentration with an exponentially weighted moving average. Ra looks like it has been trending down, while HCL concentation has different periods. Create a column for the day and summarize Ra and HCL by day. Then use Bivariate as Jim suggested. The results are below.  Note the means of the control charts and teh means of the correlation are not identical, since the means on the control charts are using all the data and teh correlation is the mean of the daily means.

 

image.pngimage.png

image.png

Benson

Occasional Contributor

Joined:

Jun 4, 2018

Re: How do I perform correlation in JMP but data is Not one-to-one correspondence?

Hi Jim, gzmorgan0

Many thanks for your answers.

1. The daily means

2. Bivariate

 

:)