Nice exbeeriment Malcolm. I occasionally like a fall seasonal beer but don’t see myself making one anytime soon, but still good to know it can be simpler than I thought
I did make pumpkin pancakes once that I thought the actual pumpkin gave a lot of flavor to. I should make those again soon, getting to be the perfect time of year.
+1
Ah man, I totally forgot about pumpkin pancakes. I had them once and, if I recall, they were pretty damn delicious.
My local makes a wildly popular pumpkin beer. One year they decided to not use the pumpkin in the recipe and they had one guy comment he really couldn’t taste the pumpkin so they put the pumpkin back in the recipe. The amount they use…drumroll…1 lb per barrel. Yes, that is right, one single pound per barrel (32 gallons) or wort. They put it in there basically to say they have it in there.
would be curious if that guy was satisfied on the new batch?
He can taste it!
[quote]Of the 40 participants, only 17 (p=0.11) were able to correctly identify the unique beer in the triangle test
[/quote]
Can anyone comment on what a typical threshold for significance is in sensory testing? Any food scientists in the house? p<0.05 seems like a pretty tough standard to me, especially with relatively small sample sizes, and several of these Xbmts have been flirting with the p<0.1 range.
For those of us who had statistics so long ago that there were no desktop computers, can you explain the symbol and the formula for this “statistical significance” (p less than .05, for example)? Near as I can tell, it comes out to about half…which leads to Sean’s question about what level is significant, whether statistically relevant or just cuz it means something that can be considered reasonably dependable to base conclusions upon.
I’ll try…
The thing that makes a triangle test so statistically useful is that it’s binary; each participant either picks correctly, or doesn’t. It’s a coin toss, and if you toss a coin enough times you can collect valid statistics about how likely you are to get various combinations of heads/tails. The probability that a given heads/tails combination will come up for a fair coin is the p-value. In this case we’re saying that, for something not to be due to chance, there has to be only a 5% probability that it occurred by chance, and therefore a 95% probability of it being an actual result. In the case of a triangle test, we’d expect people to guess correctly about 1/3 of the time, and so, as you noted, once the numbers of correct responses gets to something more like 1/2, it becomes significant. Exactly what that fraction needs to be for it to be significant depends on the desired p-value.
This is called a binomial distribution, by the way, and isn’t limited to equally weighted outcomes or true/false outcomes, they just make the math much easier.
Edit: In the hard sciences (or at least in my experience), p<0.05 is generally acceptable for publication of empirical data. p<0.01 is generally desirable and would be accepted for publication pretty much anywhere.
That was a great write up. Thank you.
I made a Pumpkin Porter last year which I roasted 3lbs of pumpkin plus added some pumpkin purée to the mash. I am personally not a big fan of the style of beer as I brewed this for my wife. However after your write up I will probably try another without the pumpkin and just use the spices. I will also not make a porter as I thought it was a little harsh.
Can anyone comment on what a typical threshold for significance is in sensory testing? Any food scientists in the house? p<0.05 seems like a pretty tough standard to me, especially with relatively small sample sizes, and several of these Xbmts have been flirting with the p<0.1 range.
[/quote]
I did quite a bit of poking around and it would seem most food scientists agree that p<0.05 is acceptable, especially using the one tailed test we recently adopted. I’ve also consulted with 4 statisticians who agree.
But, I’m neither a food scientist not a statistician. I just homebrew and ask people questions 
Thanks, Sean, I was hoping one of you smart guys could set it straight for me - too many years away from college. I remember some things, but others, not so much. P is for Probability, but here maybe the high ranges for science publication are not necessary, so you are asking if there is an industry standard for non-science probabilities that would still be considered reasonable to use to authenticate a conclusion. If I caught your drift…
Thank you, Tobby. But Marshall did have his hands involved as always. He’s a good editor!
FYP
[/quote]
No worries, Denny! It is his site so easy to mix up. Per my response to Toby - Marshall always has input, editing, and other contributions.
Oops! TOBY. I added a B.
No charge. :o
lol…I was about to FYP too, but never mind. Yeah, it’s Marshall’s site, but since you performed and wrote up the xbmt, thought you deserved to get noticed for it. Don’t want Marshall’s ego to blow up out of control either. 