커피 장비

  

BWIssue 12.10.23. 01:37
댓글 0 조회 수 2880

461_Grinder_Trio_1.jpg


모든 것이 동일한 상태에서 펌프만 다른 에스프레소 머신 2 대로 진행한 Ken Fox 와 Bob 의 테이스팅 실험 결과.

코니컬과 플랫 하이브리드 버를 사용한 심발리의 Max 와 일반 플랫버 Cadet 의 비교 테스트였으며, 굉장히 의미있고 재미있는 결과를 보여준다. 발췌한 내용은 Jim Schulman 이 이들의 결과를 통계학적으로 분석해서 정리한 결과.


HB 의 Dan 에 의해서 제의된 Titan Grinder Project 의 번외편 격인 테스트이기도 하지만, 단순히 하이브리드와 플랫의 날이 차이가 가져오는 결과 이전에 새로운 변수가 하나 더 추가되었고, 또한 이 변수가 던져주는 의미는 역시나 주목해볼만한 가치가 있다.






DATASET
----------------------------------

# MA TA PR FA CR MF
01 ro ke -2 -1 +1 -2
02 ro ke +2 -1 +2 +2
03 vi ke -2 +2 -1 -2
04 vi ke -2 -1 -2 -2
05 ro ke -2 0 +1 -1
06 vi ke +2 +2 -1 +2
07 ro bo +2 +3 +2 +2
08 ro bo +2 +3 +2 +3
09 vi bo +1 +1 +1 +1
10 vi bo -2 -3 -2 -2




Each row represents a pair of shots, one ground on the Max, one on the Cimbali Cadet (same as the Junior, with new burrs), presented to the taster simultaneously and unidentified. The numerical scores are comparative. Negative indicates a preference for the Cadet, positive for the Max. 3 is a large preference, 2 a distinct preference, 1 a mild preference. Tasters are instructed to avoid 0, no preference, if at all possible.

MA is the machine on which the Max shot was pulled: "vi" for vibe, "ro" for rotary. The other machine was used for the cadet shot.
TA is the taster for that pair of shots: "ke" for Ken, "bo" for Bob
PR is overall preference, -3 to +3, +ve means better for the Max grinder
FA is flavor/aroma strength, -3 to +3, +ve means better for the Max grinder
CR is crema appearance, -3 to +3, +ve means better for the Max grinder
MF is mouthfeel, -3 to +3, +ve means better for the Max grinder


ANOVA SUMMARIES



HOW TO READ THE SUMMARIES

The block labelled "coefficients" shows the effect of each of the variables:

-- The variable labelled "(Intercept)" shows how much the Max (+ve) or Cadet (-ve) was preferred **after** all the other variables are eliminated. This is the paydirt variable.

-- The variable labelled "mavi" shows the difference the machine made. If it is negative, Cadet shots taste better on the vibe than the rotary, if positive, Max shots taste better on the vibe than the rotary.

-- The variable labelled "take" shows the difference the taster made. If it is negative, Ken likes the Cadet more than Bob, if it is positive, Ken likes the Max more than Bob

-- I could explain the interaction term "mavi:take," but your head would explode. It's in there to remove the remaining extraneous factor from the payoff variable, the intercept. It is not meaningful by itself.

Of the columns, the "Estimate" and the Pr(>|t|) are important:

-- Estimate is the degree to which the Max is preferred. This is in the same scale as the original data.

-- Pr(>|t|) shows how often this result could occur by chance if there were no differences at all. A value less than 0.05 (5%) is mildly significant. A value less than 0.01 (1%) is very significant.

I have left out the statistics for the model as a whole, since the purpose is to get the individual variable's affects, not to predict the exact tasting outcome.




OVERALL PREFERENCE

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 2.000 1.467 1.363 0.222
mavi -2.500 2.075 -1.205 0.274
take -2.667 1.894 -1.408 0.209
mavi:take 2.500 2.679 0.933 0.387



FLAVOR/AROMA STRENGTH

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 3.000 1.106 2.714 0.0349 *
mavi -4.000 1.563 -2.558 0.0430 *
take -3.667 1.427 -2.569 0.0424 *
mavi:take 5.667 2.018 2.807 0.0309 *



CREMA APPEARANCE

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 2.0000 0.6972 2.869 0.0285 *
mavi -2.5000 0.9860 -2.535 0.0444 *
take -0.6667 0.9001 -0.741 0.4869
mavi:take -0.1667 1.2729 -0.131 0.9001



MOUTHFEEL

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 2.500 1.424 1.756 0.130
mavi -3.000 2.014 -1.490 0.187
take -2.833 1.838 -1.541 0.174
mavi:take 2.667 2.600 1.026 0.345

ANALYSIS

The Max did a better job in all the categories. For the significance of the result see below.

There is a big surprise in the "mavi" variable; add or subtract it from the intercept term to see what I mean. In essence, the Max spanked the Cadet when it was paired with the rotary, and did slightly worse when paired with the vibe. Bob liked the Max a bit more than Ken did. The direction of the coefficients are consistent across all four of the taste attributes.

This outcome is consistent across all the taste categories. This consistency adds to the significance of the result, since if the result were just chance, it would be like flipping 4 heads in a row. This can be seen when all 4 taste categories are used in a single analysis



A: Done so that there are 40 data points:

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 2.3750 0.5577 4.258 0.000141 ***
mavi -3.0000 0.7887 -3.804 0.000533 ***
take -2.4583 0.7200 -3.414 0.001598 **
mavi:take 2.6667 1.0183 2.619 0.012829 *


B: Done with 10 data points, scores added together and divided by 4

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 2.375 1.032 2.302 0.0609 .
mavi -3.000 1.459 -2.056 0.0855 .
take -2.458 1.332 -1.846 0.1144
mavi:take 2.667 1.883 1.416 0.2066


CONCLUSION

My conclusion is that the Max grinds an immediately perceivable better shot on the rotary Cimbali than the Cadet or Junior grinder. On the Vibe machine, the edge disappears completely. Whether this conclusion about the Max applies to all rotary and vibe machines is unknown. However, the results warn that one needs to check for an interaction between conical grind quality and the pump being used. 

profile

ABOUT ME

대한민국 No.1 커피 미디어 "블랙워터이슈" 입니다.
일반회원
Where Espresso Machines Are Born
Slayer and pre-brew
Bonavita versus Technivorm
Faema President brewing temp testing
vacuum breaker 의 역할