Fit Model Standard Least Squares - Differences between JMP and Excel Linest function

[#1]

I'm working with some data and trying to fit a multiple linear regression model, but I've noticed parameter differences between using the "Linest" function of Excel, and the Fit Model - Standard Least Squares function of JMP17.2. The parameters of the model (intercept and each X variable coefficeint) are close to each other, but not identical. The in-going data is identical.

How does the linest function of Excel differ to the way least squares is applied in JMP 17.2? Which is 'better'? Or how are they different? Is it possible to make JMP behave exactly the same as excel, or vice versa?

[#2]

Hi @ScrivensL ,

  You might need to share some data (anonymize it, of course), or explain a little bit more what the problem is, because when I compare the SLS Fit Model from JMP and the LINEST function in Excel, I get the exact same answers from both.

  As an example, I am using data from MS, you can find it here from the definition page of LINEST. When you run the SLS Fit Model in JMP, you get this as the summary report:

  If you do the same multiple regression in Excel, you get this for the output with the same input data as with JMP:

  The bottom portion of the screenshot shows how the data from the LINEST function is laid out when turning on the stats option in the function. If you look closely, you'll see that the estimates for the fit as well as the statistics are all identical to JMP. The only thing that shows up as being different is the rounding, but that's because JMP's default is set to a smaller number of significant digits than Excel.

  So to me, I'm not sure where you are seeing a discrepancy. I'd have to use your same data to see if I can reproduce the differences that you're seeing. With this data, I get the same from both software.

Hope this helps!,

DS