Multivariate correlation?

Can anyone see what I am doing wrong or otherwise explain how to calculate the multivariate correlation or Rsq using Excel formulas, not the Data Analysis Regression tool?

I confess that I don't know what I'm doing.  I am not familiar with multivariate regression analysis.

Refer to the example at http://www.ehow.com/how_6049413_use-correlation-data-three-sensors.html.

When I follow those instructions, I get the following results from the Data Analysis Regression tool.

So the (adjusted) multivariate Rsq is about -0.2176.

(What "adjustment"?)

But according to http://en.wikipedia.org/wiki/Multiple_correlation, we can compute the multivariate Rsq as follows:

I believe I implemented that correctly as follows:

Note that the calculated Rsq is about 0.0868.  Not only is that not even close to "adjusted Rsq", but also it is a different sign.

Aha!  How can (adjusted) Rsq be negative in the first place?  R, yes; Rsq, no.  Well, maybe the operative word, again, is "adjusted".

Anyway, the point is:  which one is correct?  Or more to the point:  how can we calculate the multivariate correlation factor R or Rsq?

(I would prefer to calculate R, not Rsq, since the sign identifies negative v. positive correlation, if that even makes sense for multivariate regression analysis.)

[EDIT] Aha!  I just noticed that the Data Analysis Rsq (not the adjusted Rsq) does match my calculation.  So perhaps the problem is not with either one, but with the instructions at www.ehow.com or my interpretation of "adjusted Rsq".  Or perhaps the "adjusted Rsq" is intended to be R (a misnomer?).  But note that sqrt of Rsq is about 0.2946 ("multiple R"), not (minus) 0.2176.  Any ideas?