Quote:
Originally Posted by avandonk
The point the article tried to make that even a 'trained' palate is open to error. This error is larger than the actual real 'differences' detected by these 'trained' palates.
We are dealing with another classic signal to noise problem. It is immaterial what the quality of the wine is meant to have, when the measurement criteria cannot accurately and consistently differentiate between these so called qualities!
As any scientist will tell you if you are looking for a signal in the noise then you may as well flip coins!
To get some sort of statistical measurement, very many 'trained' palates should test the wines in question until a two sigma or a 95% probabilty is reached of all the measurements. Then the mean or median of these measurements could then be considered being close to the real 'value'.
My feeling is that to reach this level of scientific consensus is that the whole vintage needs to be consumed in order to test it!!
Bert
|
This is why scientific research in the field of sensory science will use an expert
panel to determine links between sensory perception and physical/chemical properties. Discrimination tests such as triangle test or duo-trio test are good for this, however for a purpose trained panel descriptive analysis is also used with an appropriately designed experiment, where panelists will rate different flavours/aromas (eg citrus, toasty, mineral etc) on a 0-10 point scale. Principle-component analysis is commonly used here to draw out correlations.
http://en.wikipedia.org/wiki/Sensory_analysis
http://en.wikipedia.org/wiki/Discrimination_testing
https://en.wikipedia.org/wiki/Princi...onent_analysis