Hi Andy,

You know me, so you know I was up all night trying to get a handle. The handle is fairly straightforward.

The variables you control (the causes) are grind, dose, basket, and shot time. The outcomes (the effects) are shot weight and post shot puck weight.

There's no problem when you use an outcome on the Y axis and a cause on the X; or use the same set up in statistical tests and regressions. Once you start mixing this up, or using ratios of inputs and outputs, outliers and other residuals start behaving in linked and funny ways. The apparent relations can change in strength wildly as the models are very slightly changed. This flaw applies to all the analyses so far, even the rank one, since we were using shot weight as a cause, and a lot of ratios of input and output variables.

Unfortunately, using the raw data, rather than ratios, gets lousy models. I need to do more work.

another_jim wrote:Hi Andy,

You know me, so you know I was up all night trying to get a handle. The handle is fairly straightforward.

The variables you control (the causes) are grind, dose, basket, and shot time. The outcomes (the effects) are shot weight and post shot puck weight.

Fortunately it sounds like it was a satisfying night for you rather than an annoying night. Of course, let me know if there is any more data I can collect. And good luck!

AndyS wrote:Fortunately it sounds like it was a satisfying night for you rather than an annoying night. Of course, let me know if there is any more data I can collect. And good luck!

I've tracked down the problem: we're making good shots.



The graph has shot time on the X, shot weight divided by dose on the Y -- I call the graph the "al effect," since we are drawing out the more concentrated shots, cutting the less concentrated ones short. The result is that the points are mostly in the lower left triangle; and the upper right one is mostly empty. This is for my data; but you'll probably see it on yours too.

This selectivity on on stopping shots messes up the statistics and the physical relations. Long shot times should produce higher solubles yields, higher weight shots, and less concentrated shots. By using al's rule (or the blonding rule), we stretch out concentrated, low flow shots and cut watery, high flow ones short. In my dataset, this has completely removed the causal relation of shot time. I'm guessing you have a similar situation.

Despite messing with the analysis, this could be a good thing. Afterall, we are looking for effects in shots made by competent baristas.

I've been watching with interest since your earlier thread about lower dosing. Your hard work has spurred us to experiment and find a better tasting cup for the effort. Just wanted to say thanks for sharing your work. You are one seriously obsessive guy, please stay off the meds and keep on pumping out the data! Will

Thanks very much for your work, I get a little panicked (and lost) with the data but I still enjoyed reading the papers and am excited to have new things to try. I had it in my head that grinding too fine meant over extraction and that too little coffee in the basket was just a bad idea. I haven't made a better shot yet, but some OK ones (after a string of lousy ones with my regular way of doing things) by trying a finer grind / smaller dose. I'm no researcher but it has been exciting and fun, which is mostly what I'm after.

Thanks,
Phil

So Abe Carmeli got a chance to taste some of the ultra-lite roasted, downdosed espressos. He agreed they were tasting in the high 4s and low 5s. Body was another story, 3ish, even for the ristretto pours. The ultra-lite roasts simply don't have that much body, even if solids in the cup are kept high.

He's more of an acid head than me, preferring dosing around a 1/2 gram higher.

another_jim wrote:So Abe Carmeli got a chance to taste some of the ultra-lite roasted, downdosed espressos. He agreed they were tasting in the high 4s and low 5s. Body was another story, 3ish, even for the ristretto pours. The ultra-lite roasts simply don't have that much body, even if solids in the cup are kept high.

That I agree with that observation. With the light-roasted, downdosed shots, the flavors can be balanced but "there's no there there" (apologies to Gertrude Stein).

I may have missed this in my brief review of the paper, but how does shot volume relate to all this? E.g. I using a small-group lever machine (45-mm Ponte Vecchio), which requires two pulls to get a 1-oz ristretto. I have been used to dosing very full (15g) but am now determined to try down-dowsing, since I like the sweeter flavors. My first test confirmed the possibilities - the shot was sweeter.

In your conclusions you discuss the tendencies for Italians to dose less for practical reasons (btw I think you should move this up to the Intro). I am correct in assuming they are dosing 13 g for full 2-oz. doppios? So how do they get this much volume without blonding? I doubt I could get a 13g, 2-oz shot without serious blonding on my lever machine.

PC

peacecup wrote:I may have missed this in my brief review of the paper, but how does shot volume relate to all this? E.g. I using a small-group lever machine (45-mm Ponte Vecchio), which requires two pulls to get a 1-oz ristretto. I have been used to dosing very full (15g) but am now determined to try down-dowsing, since I like the sweeter flavors. My first test confirmed the possibilities - the shot was sweeter.

im not an authority on this, but from what i read, the entire paper focuses on it..

peacecup wrote:In your conclusions you discuss the tendencies for Italians to dose less for practical reasons (btw I think you should move this up to the Intro). I am correct in assuming they are dosing 13 g for full 2-oz. doppios? So how do they get this much volume without blonding? I doubt I could get a 13g, 2-oz shot without serious blonding on my lever machine.
PC

this is also covered in the paper, the 2nd link in the abstract:
http://users.ameritech.net/jim_sc..._extraction.htm#_b

and in a previous thread by Jim:
http://www.home-barista.com/forum...ference-t3182.html

Ok. One of the things I was assuming when I started the research is that the perception of the coffee's body and strength mostly depended on how much coffee solubles were in how much water.

This is not quite Andy's brew ratio, since soluble yields figures in -- it's the puck's weight loss divided by the shot weight. When I did the research I kept this figure fairly constant regardless of the dose or solubles yield. This is very easy -- if you stop the shot at the same blonding point each time, regardless of how long it takes, you'll have roughly the same solids concentration in the cup.

Turns out this is probably wrong. If the solids concentration is kept the same, underextracted shots taste stronger than properly extracted ones -- the caramels have a mellowing effect. I happen to prefer hi-powered coffees that are undrinkable when underextracted, mellowed out by deeper extractions; but this is not everyone's cup of tea. Medium and medium dark roasts have more body than light roasts, even at the same solid concentration. I do not know the chemistry and organoleptics behind this.