formula R square

[meeihualin]

Version: 2008 Operating System: Mac OS X 10.5 (Leopard) Processor: Intel I was trying to get R square (squared correlation coefficient) values of my data using the formula RSQ. However, sometimes it gave R values rather than R square values. This problem seemed to be "contagious". Whenever this problem appeared on one file while another good file was opened, all the R square values in the good file were converted to R value by themselves. Please instruct how to solve this annoying problem. Thank you!

 

[Mike Middleton]

meeihualin -

Your problem was discussed in a sixteen-message thread during
August-November 2009, and I don't think there was ever any resolution.

One way to view that thread is to browse to http://groups.google.com, click
the Advanced Groups Search link, for Group type
microsoft.public.mac.office.excel, for Subject type "RSQ function" (without
the quotes), and click Advanced Search.

Many Mac users, including me, cannot replicate your problem. To sort out the
cause, it might help if you specify your exact version of Excel. I am using
Mac Excel 2008 version 12.2.4 and Mac Excel 2004 version 11.5.8.

A workaround is to use =CORREL(...,...)^2 or =PEARSON(...,...)^2.

- Mike
http://www.MikeMiddleton.com



Version: 2008 Operating System: Mac OS X 10.5 (Leopard) Processor: Intel I
was trying to get R square (squared correlation coefficient) values of my
data using the formula RSQ. However, sometimes it gave R values rather than
R square values. This problem seemed to be "contagious". Whenever this
problem appeared on one file while another good file was opened, all the R
square values in the good file were converted to R value by themselves.
Please instruct how to solve this annoying problem. Thank you!

 

[XinXin]

Hi Mike, if you can find what's wrong, and it turns out to be a product debug, please do let me know ([email protected]). We will investigate futher here at Microsoft.

Thanks,
XinXin Liu
Test Lead, Macintosh Business Unit, Microsoft

 

[Carl Witthoft]

I'll send you a sample workbook, but it's easy enough to do:

just build a table of y = a*b^x +rand() and use logest() to get the
coefficients as well as the additional statistics.
Then calculate the fitted y' values using the logest values for a and
b, and calculate the sum ((y-y')^2)

Carl