[tx_eggman]

exp(aj + bj xi ) / ∑k=1J exp(ak + bk xi ) , (1)

This is where they lost me ...

[Robert A Cook PE]

Look at the conclusions and analysis, not the math. (If you trust their math, that is.)

From their exclusions (what data was selected and why, and what data was rejected and why) and conclusions, you can judge their accuracy.