Jump to content


Photo
- - - - -

Flaked vs No-Flaked


  • Please log in to reply
34 replies to this topic

#21 3rd party JKor

3rd party JKor

    Puller of Meats

  • Patron
  • PipPipPipPipPipPip
  • 63966 posts
  • LocationNW of Boston

Posted 22 November 2016 - 09:16 AM

Yes, what they're doing produces statistically significant results but n is still 1.

#22 denny

denny

    Living Legend

  • Members
  • PipPipPipPip
  • 9090 posts
  • LocationEugene OR

Posted 22 November 2016 - 09:34 AM

Have they actually had an experiment that showed significant results yet? I feel like every one that i've read shows that there's no difference between process/ingredient A and process/ingredient B.

Maybe it's time they do one that is a slam dunk just to validate their methodology.

 

At EB, we've only had a couple that showed definite significant results.  That's just the way it works with citizen science.  We hope that our larger sample size will give us more definitive results, but truthfully it's about the same as Marshall.



#23 denny

denny

    Living Legend

  • Members
  • PipPipPipPip
  • 9090 posts
  • LocationEugene OR

Posted 22 November 2016 - 09:37 AM

yeah.  that fermentor one really makes me wonder if there is something else going on that is messing up their experiment.  at our scale how much of a difference can those geometries and time spent in a container that is slightly more or less oxygen permeable really matter?

 

As much as I love Marshall his guys and respect what they do, one problem is that they draw their results from only a single trial.  That's why we have IGORs.  We want to make sure that every experiment has multiple testers and a larger pool of tasters.


The strangest thing was that for the PET vs Glass, even Marshall couldn't tell the difference in his own blind triangle tests, but the tasters could tell the difference. It seems like in almost every experiment the experimenter is able to pick out the different beers even when the tasters don't show significant results. Having it go the other way is curious.

I thinknat some point you also have to consider that every fermentation is different regardless of how much you control. I'd be curious to see results from two different fermenters of the same material. I.e. Two identical PET carboys.

 

Again, why we use multiple testers.



#24 3rd party JKor

3rd party JKor

    Puller of Meats

  • Patron
  • PipPipPipPipPipPip
  • 63966 posts
  • LocationNW of Boston

Posted 22 November 2016 - 09:47 AM

Ideally you'd have 10-30 sample pairs (each from a different fermentation trial) and each would have an independent set of 20-30 testers. Of course, on the upper end you're talking about nearly 1,000 independent testers. It's either unrealistic or a good starting point for a peer reviewed article in a technical journal. :D

Edited by JKor, 22 November 2016 - 09:47 AM.


#25 denny

denny

    Living Legend

  • Members
  • PipPipPipPip
  • 9090 posts
  • LocationEugene OR

Posted 22 November 2016 - 10:12 AM

Ideally you'd have 10-30 sample pairs (each from a different fermentation trial) and each would have an independent set of 20-30 testers. Of course, on the upper end you're talking about nearly 1,000 independent testers. It's either unrealistic or a good starting point for a peer reviewed article in a technical journal. :D

 

That's what we're shooting for in the IBU experiment currently underway.  16-18 brewers, likely 25-30 different samples.  Hopefully each of those brewers can get at least 10 tasters.  Plus all beers will be lab analyzed.



#26 3rd party JKor

3rd party JKor

    Puller of Meats

  • Patron
  • PipPipPipPipPipPip
  • 63966 posts
  • LocationNW of Boston

Posted 22 November 2016 - 11:13 AM

That's what we're shooting for in the IBU experiment currently underway. 16-18 brewers, likely 25-30 different samples. Hopefully each of those brewers can get at least 10 tasters. Plus all beers will be lab analyzed.


:frank:

#27 Brauer

Brauer

    Frequent Member

  • Members
  • PipPipPipPip
  • 1857 posts
  • Location1 mile north of Boston

Posted 22 November 2016 - 11:26 AM

The fermentor differences may be an example that shows that p≤0.05 is not a particularly stringent standard, so you'll still find occasional differences by pure chance.

Biological science is usually performed with 3 or (extravagantly) 5 replicates, not dozens or thousands, unless you are trying to isolate an expectantly rare occurrance. As brewers, we aren't interested in rare events, just typical ones. This is often made up for by repeating the experiment ~3 times, to make sure that the results are consistent, but if a result is significant, it shouldn't be so elusive.

With a lot of repeat runs, you do run the risk of finding a difference because of an oulier set (like someone contaminating a batch) that skews the results. You can also run into the situation where the quirks of statistics say that no individual experiment finds a difference, but the data on the whole show a difference just because of the larger sample size. This might be corrected by abandoning the rather generous p≤0.05 for a more stringent p≤0.01 to help correct for the looseness of the experimental design.

These are some reasons that statistical relevence is usually determined from a single experiment, not a combination of a series of experiments. Instead, you might compare the number of experiments that showed a difference to the number that didn't and determine if that difference is significant. For example, in the situation where one batch was dramatically different due to a contamination, the combined numbers could indicate a difference, but the fact that only 1 out of 5 sets was statistically different would lead to a conclusion that it probably was a fluke.

#28 denny

denny

    Living Legend

  • Members
  • PipPipPipPip
  • 9090 posts
  • LocationEugene OR

Posted 22 November 2016 - 12:10 PM

The fermentor differences may be an example that shows that p≤0.05 is not a particularly stringent standard, so you'll still find occasional differences by pure chance.

Biological science is usually performed with 3 or (extravagantly) 5 replicates, not dozens or thousands, unless you are trying to isolate an expectantly rare occurrance. As brewers, we aren't interested in rare events, just typical ones. This is often made up for by repeating the experiment ~3 times, to make sure that the results are consistent, but if a result is significant, it shouldn't be so elusive.

With a lot of repeat runs, you do run the risk of finding a difference because of an oulier set (like someone contaminating a batch) that skews the results. You can also run into the situation where the quirks of statistics say that no individual experiment finds a difference, but the data on the whole show a difference just because of the larger sample size. This might be corrected by abandoning the rather generous p≤0.05 for a more stringent p≤0.01 to help correct for the looseness of the experimental design.

These are some reasons that statistical relevence is usually determined from a single experiment, not a combination of a series of experiments. Instead, you might compare the number of experiments that showed a difference to the number that didn't and determine if that difference is significant. For example, in the situation where one batch was dramatically different due to a contamination, the combined numbers could indicate a difference, but the fact that only 1 out of 5 sets was statistically different would lead to a conclusion that it probably was a fluke.

 

We've spent a lot of time discussing outliers and how to deal with them.



#29 Brauer

Brauer

    Frequent Member

  • Members
  • PipPipPipPip
  • 1857 posts
  • Location1 mile north of Boston

Posted 28 November 2016 - 06:48 AM

The efficiency was pretty low, overall, and lower in the flaked oats batch, indicating that a lot of potential contribution from the oats was left behind. Also, I'd like to have seen the comparison made with cooked oats. The consistancy of oats brought to a boil is so different than oats soaked in 150°F water, I have to wonder if cooking is required to solubilize their beta-glucans.

#30 positiveContact

positiveContact

    Anti-Brag Queen

  • Patron
  • PipPipPipPipPipPip
  • 68886 posts
  • LocationLimbo

Posted 28 November 2016 - 07:12 AM

Are flaked oats different than quick oats?

#31 Brauer

Brauer

    Frequent Member

  • Members
  • PipPipPipPip
  • 1857 posts
  • Location1 mile north of Boston

Posted 28 November 2016 - 07:35 AM

Are flaked oats different than quick oats?

Quick oats are usually chopped finer than regular and flaked oats.

#32 Brauer

Brauer

    Frequent Member

  • Members
  • PipPipPipPip
  • 1857 posts
  • Location1 mile north of Boston

Posted 28 November 2016 - 08:31 AM

Quick oats are usually chopped finer than regular and flaked oats.

The quick oats I've seen have looked chopped. The description I found said quick oats were rolled thinner. By "regular oats", I mean rolled oats, which are the same as flaked oats.

#33 mabrungard

mabrungard

    Advanced Member

  • Members
  • PipPipPip
  • 109 posts
  • LocationIndianapolis, IN

Posted 29 November 2016 - 07:46 PM

I'm with Denny on the mythology of oats. I don't find that they are better at adding body or anything. And I recall Professor Michael Lewis finding that oats in a stout just made the beer more astringent. 

 

I found that flaked barley adds a huge amount of head formers to beer, but I can taste the barley in pale beers. Flaked wheat also adds head formers, but at a much lower rate than flaked barley. But I find that the flavor imparted by the flaked wheat is much more pleasant and pleasing than barley. By the way, the amount of beta-glucan in barley is about 10 times greater than wheat. That is probably the difference in the head building ability.



#34 neddles

neddles

    No Life

  • Patron
  • PipPipPipPipPip
  • 16518 posts

Posted 29 November 2016 - 08:24 PM

https://scottjanish....e-brewing-oats/

 

maybe we aren't using enough?



#35 Brauer

Brauer

    Frequent Member

  • Members
  • PipPipPipPip
  • 1857 posts
  • Location1 mile north of Boston

Posted 30 November 2016 - 06:53 AM

Here's a great review on Oats by Schnitzenbaumer.

It's odd that Lewis would have found increased astringency from Oats, since they appear to contribute less polyphenols than Barley.


0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users