Years ago I brewed my own beer, and things weren't _too_ complicated back then, but I quit for awhile and got involved in video poker, and then all of this math stuff just seemed to come from no where, ... not that I really mind _too_ much since I'm about _13%_ nerd. Well, it started to give me a headache anyway, so I decided to go back to homebrewing for a change of pace -- to relax my mind a bit -- and what do I find on rec.crafts.brewing? THIS (below)! Steve Jacobs and Harry Porter, et al ... see what you've done?! Now you guys are going to have to help me solve this. 
Bill Velek
P.S. My next barley wine is going to be called "Jerks or Better 9.6" ... meaning 9.6% alcohol
···
*******
So, the only statistically correct model is:
lm(formula = attenuation ~ sg, data = mashdata)
Residuals:
Min 1Q Median 3Q Max
-4.6437 -2.6291 -0.5675 1.4207 8.0444Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 92.85068 5.27557 17.60 2.79e-08 ***
sg -0.32355 0.09459 -3.42 0.00762 **
---
Signif. codes: 0 `***' 0.001 `**' 0.01 `*' 0.05 `.' 0.1 ` ' 1Residual standard error: 4.029 on 9 degrees of freedom
Multiple R-Squared: 0.5652, Adjusted R-squared: 0.5169
F-statistic: 11.7 on 1 and 9 DF, p-value: 0.007623I repeated the analysis with real attenuation and it was essentially the same. The correlation between real and apparent attenuation is 0.99891095, which is probably just roundoff error. Here's the linear model with real attenuation:
lm(formula = ra ~ sg, data = mashdata)
Residuals:
Min 1Q Median 3Q Max
-3.7071 -2.0288 -0.3061 1.3078 6.4434Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 75.82589 4.22043 17.966 2.33e-08 ***
sg -0.27217 0.07567 -3.597 0.00578 **
---
Signif. codes: 0 `***' 0.001 `**' 0.01 `*' 0.05 `.' 0.1 ` ' 1Residual standard error: 3.223 on 9 degrees of freedom
Multiple R-Squared: 0.5897, Adjusted R-squared: 0.5441
F-statistic: 12.94 on 1 and 9 DF, p-value: 0.005779