Question-and-Answer Resource for the Building Energy Modeling Community
 | 1 | initial version |
After doing some testing on this spreadsheet today and then coming across this post, I think the main issue is that it is still calculating coefficients for linearly dependent (constant across the data set) variables, as opposed to leaving those variables out of the matrix calculations and setting the related coefficients to 0 as Richard said above (which is what should happen). This, in turn, causes all of the coefficients be erroneous since they are solved for simultaneously.
So what this means, as Richard implies, is that this spreadsheet is limited to scenarios where data is available with variation (linearly independent) in all relevant variables, which are: Qtot - Twb, Tw, Vair, Vw; Qsens - Tdb, Twb, Tw, Vair, Vw; Power - Twb, Tw, Vair, Vw. Therefore, if any of your variables are linearly dependent, the coefficient sets that are produced will not be realistic, as the OP saw in error comparison.
