Statistical significance of mash pH estimates?

[sixhotdogneck]

1.) Water/Mash calculators attempt to calculate the mash pH at chemical equilibrium.
2.) Homebrewers are incorrectly measuring the pH before chemical equilibrium.
3.) Homebrewers are incorrectly using their pH meters.
4.) Homebrewers do not measure the chemical equilibrium pH they simply "trust" the calculator.
5.) Homebrewers claim their water/mash calculators help them make good beer.

****************************************************************************

1.) What would the process be to verify the estimate of each water/mash calculator?
2.) How many samples would be required to make verification statistically significant?
3.) How close would the estimate have to be to the actual to be considered a good estimate?

 

[corax]

Or here's a scary thought: maybe being +/- 0.2 pH units off your target doesn't make squat difference.

 

[mongoose33]

1.) Water/Mash calculators attempt to calculate the mash pH at chemical equilibrium.
2.) Homebrewers are incorrectly measuring the pH before chemical equilibrium.
3.) Homebrewers are incorrectly using their pH meters.
4.) Homebrewers do not measure the chemical equilibrium pH they simply "trust" the calculator.
5.) Homebrewers claim their water/mash calculators help them make good beer.

****************************************************************************

1.) What would the process be to verify the estimate of each water/mash calculator?
2.) How many samples would be required to make verification statistically significant?
3.) How close would the estimate have to be to the actual to be considered a good estimate?

I agree with Corax, you're measuring cordwood with a micrometer (I read that somewhere, liked it).

To answer:

1. You'd have to do a lot of different malts, water profiles, etc. to do that. How you'd do that I have no idea, and I'm certainly not volunteering to help.

2. It's not how many samples, it's how many instances or cases. I can tell by the question that perhaps there may be some value in your taking a statistics course.

3. Again, you need to take a statistics course. How close depends on the standard deviation, the number of cases, and your own inclination as to how close is close enough. There is no standard against which you can judge this.

IMO, you're tilting at windmills here.

 

[Blazinlow86]

1.) Water/Mash calculators attempt to calculate the mash pH at chemical equilibrium.
2.) Homebrewers are incorrectly measuring the pH before chemical equilibrium.
3.) Homebrewers are incorrectly using their pH meters.
4.) Homebrewers do not measure the chemical equilibrium pH they simply "trust" the calculator.
5.) Homebrewers claim their water/mash calculators help them make good beer.

****************************************************************************

1.) What would the process be to verify the estimate of each water/mash calculator?
2.) How many samples would be required to make verification statistically significant?
3.) How close would the estimate have to be to the actual to be considered a good estimate?

I'm not sure I understand your question? I use brunwater to get a estimated amount of lactic acid too add. After doing so I verifiy it's correct with my meter. If it's off I add more acid until I reach my target. I've found brunwater to be basically close enough within a few points. I've also tried beersmiths tool however its always recommend basically double what brunwater does. Again because I'm checking with my meter everytime/multiple times I know adding beersmiths suggestions would bring the ph way below my target and out of range. Cheers

 

[Blazinlow86]

Or here's a scary thought: maybe being +/- 0.2 pH units off your target doesn't make squat difference.

I agree with this comment. I don't think it could really be debated that ph makes a difference however I think it's more in the range than exact that matters but It's also all somewhat subjective. I brewed a terribly over astringent IPA once and my buddy LOVED it. I couldn't even drink it. cheers

 

[sixhotdogneck]

A lot of rude and unnecessary negativity, insults and angst surrounding these questions.

Imagine what could be accomplished if you changed your attitudes and helped establish a procedure to test the calculators.

Say by following the procedures we began to collect data and discovered that Brunwater was consistently .2 pH off or that MME was always 1.5pH high, we could then use the context of the recipes and the data to further enhance the calculators, right?

 

[sixhotdogneck]

Or here's a scary thought: maybe being +/- 0.2 pH units off your target doesn't make squat difference.

That's exactly what I said in my post, people are making good beer even though not everything is measured/estimated accurate or done "properly".

 

[sixhotdogneck]

I agree with Corax, you're measuring cordwood with a micrometer (I read that somewhere, liked it).

To answer:

1. You'd have to do a lot of different malts, water profiles, etc. to do that. How you'd do that I have no idea, and I'm certainly not volunteering to help.

2. It's not how many samples, it's how many instances or cases. I can tell by the question that perhaps there may be some value in your taking a statistics course.

3. Again, you need to take a statistics course. How close depends on the standard deviation, the number of cases, and your own inclination as to how close is close enough. There is no standard against which you can judge this.

IMO, you're tilting at windmills here.

Your post is rude and insulting. Just because I don't have a college education in statistics doesn't mean I can't contribute ideas.

 

[Blazinlow86]

A lot of rude and unnecessary negativity, insults and angst surrounding these questions.

Imagine what could be accomplished if you changed your attitudes and helped establish a procedure to test the calculators.

Say by following the procedures we began to collect data and discovered that Brunwater was consistently .2 pH off or that MME was always 1.5pH high, we could then use the context of the recipes and the data to further enhance the calculators, right?

I'm not sure there is much more that can be done at this point. We already have a procedure to see how far off these programs are by just using them and then checking with the meter. I don't think your ever going to get it 100% as there's too many variables and imho who cares as .03 off is good enough. Mme is advertised as being superior by the creator *if* you take ph readings in advance of all your grain but I can't confirm that as it's way too complicated to possibly be worth it for .03 accuracy. I did try Mme *without* using all the extra stuff and it was way off similar to bs3. I could have been using it incorrectly however. Cheers

 

[mongoose33]

Your post is rude and insulting. Just because I don't have a college education in statistics doesn't mean I can't contribute ideas.

Perhaps you can tell me how I can explain this to you in a style to which you will not take offense?

My point is that what you want to do you can't do. How would you like me to deliver that message?

 

[sixhotdogneck]

So if 100 mashes are measured you have 100 samples (you're using the term "case" or "instance").

Each sample is mated with an estimate from each of the 5 calculators for that recipe.

So 100 samples - 5 estimates for each sample.

What could be derived from that?

 

[sixhotdogneck]

Perhaps you can tell me how I can explain this to you in a style to which you will not take offense?

My point is that what you want to do you can't do. How would you like me to deliver that message?

Telling people who are asking questions about statistics to take a statistics course is rude and makes you look like an a*shole.

 

[mongoose33]

Telling people who are asking questions about statistics to take a statistics course is rude and makes you look like an a*shole.

Again, how should I have said this? I'm sorry you felt offended, but should I just have excised the thing about stat courses? Is that what the issue is?

You're asking questions that cannot be answered simply; unless you want me to deliver a statistics course here on HBT, how else could that information be delivered to you? An answer involves a normal curve, standard deviation, what a standard deviation represents, sufficient sample size, probability, confidence intervals, and signficance level and what that means. And if you want to posit a standard against which you'd develop an alternative hypothesis, then hypothesis testing.

And FWIW, I do teach statistics at the college level. You're asking "statistics" to do something it cannot do. If the questions were different "statistics" might be relevant, but not in this case.

 

[sixhotdogneck]

I do teach statistics at the college level.

That explains your bad, know it all, holier than thou attitude...

Instead of insulting and throwing around statistics terms why not take a step down and in plain English explain what you believe would need to be done.

 

[Blazinlow86]

I'm still unsure what your direct question is? Are you trying to find out what program works best without a meter? It's pretty widely agreed on that brewnwater is generally the closest and the go to. In my experience it's been in the range everytime. I check with a better then normal meter each time. I can't imagine it being closer realistically? Side note I haven't read anything here that's anything insulting. Maybe speaking bluntly but definitely nothing off side

 

[RPh_Guy]

2.) Homebrewers are incorrectly measuring the pH before chemical equilibrium.
3.) Homebrewers are incorrectly using their pH meters.
4.) Homebrewers do not measure the chemical equilibrium pH they simply "trust" the calculator.
5.) Homebrewers claim their water/mash calculators help them make good beer.

A lot of rude and unnecessary negativity, insults and angst surrounding these questions.

Hmm I wonder why, when you start with generalizations that [all] homebrewers are doing it all wrong.

Are you making the One True Beer by chance?

Maybe it's time to have a homebrew and relax a little bit. We're making beer, not writing a holy text.

 

[sixhotdogneck]

Hmm I wonder why, when you start with generalizations that [all] homebrewers are doing it all wrong.

Are you making the One True Beer by chance?

Maybe it's time to have a homebrew and relax a little bit. We're making beer, not writing a holy text.

Have you read the other threads?

https://www.homebrewtalk.com/forum/...ablished-for-mash-ph-measurement-time.654426/
https://www.homebrewtalk.com/forum/...their-ph-meters-properly.654439/#post-8373586

 

[ajdelange]

All the statisticians in the world laid end to end wouldn't reach.....

ready for this?

.... a conclusion!

 

[sixhotdogneck]

Are you trying to find out what program works best without a meter?

No. Gathering statistics about the deviation from the estimate might lead to clues as to how these estimation programs could improve.

 

[Biscuits]

I'm still unsure what your direct question is? Are you trying to find out what program works best without a meter? It's pretty widely agreed on that brewnwater is generally the closest and the go to. In my experience it's been in the range everytime. I check with a better then normal meter each time. I can't imagine it being closer realistically? Side note I haven't read anything here that's anything insulting. Maybe speaking bluntly but definitely nothing off side

I think his question is more general. If I understand it correctly, I think he is asking whether or not we can prove that mash pH software is actually as accurate as we believe it to be. With all of the variables that come along with that, which he mentioned in his original post, what path can we take to form a solid foundation to prove or disprove the existing calculators.Personally, I do not know if my beer has gotten better because of how I treat my water...the jury is still out on whether it is the water or whether I have simply gotten more comfortable with my process. What I do know, is that brewing is science...and water is just a small peice of the puzzle when it comes to making high quality beer....So, in my mind, it's necessary to learn, but only a drop in the ocean in regards to what you need to know to be at the level of <Insert your favorite craft brewery here>.

 

[Silver_Is_Money]

Another thread has relevance here. Any mash pH sample taken prior to the 30 minute mark is likely to be a flawed measurement. Here is the link to the thread:

https://www.homebrewtalk.com/forum/...ablished-for-mash-ph-measurement-time.654426/

 

[Blazinlow86]

I think his question is more general. If I understand it correctly, I think he is asking whether or not we can prove that mash pH software is actually as accurate as we believe it to be. With all of the variables that come along with that, which he mentioned in his original post, what path can we take to form a solid foundation to prove or disprove the existing calculators.Personally, I do not know if my beer has gotten better because of how I treat my water...the jury is still out on whether it is the water or whether I have simply gotten more comfortable with my process. What I do know, is that brewing is science...and water is just a small peice of the puzzle when it comes to making high quality beer....So, in my mind, it's necessary to learn, but only a drop in the ocean in regards to what you need to know to be at the level of <Insert your favorite craft brewery here>.

Hmm not trying to beat a dead horse here but I'm still not understanding. We can prove how accurate the software is by using at and taking the actual ph readings to compare. I do it every brew . Unless my meter is incorrect and I calibrate it before use what are we trying to figure out? In 55+ batches brunwater has never been off by .04. is the question specifically directed at people without meters trying to verify if it's correct?

 

[ajdelange]

We know the answer to that. There are two parts to estimating mash pH. Accurately modeling the titration (acid/base) properties of malts so that the calculator has a representation of the malt the brewer is actually using is the first part. The second part is correctly processing the data represented by those models and, in that processimg accurately interpolating water, lactic acid, phosphoric acid... which may be added in an attempt to control mash pH. The current calculators attempt to do both parts and do neither well. There are some real, in fact, I think, unsurmountable, problems with part 1. Unless you measure each malt you intend to use and do it properly (which is a very laborious process described in a Sticky) you must rely on measurements made by someone else at some other time on some malt sample that is not the one you have in the sack on your brewery floor. The only reasonable solution I see to this problem is to have the maltsters undertake the measurements as part of their evaluation of each lot number. They take one measurement of pH (congress mash), why can't they take a couple more? Don't know but they can't.

The second part can be addressed (and has been on the prototype level). What I call the Gen I calculators commit a variety of errors with respect to how they handle alkalinity, how they treat malt buffering and how, ultimately they solve for pH. We now how to correct all these errors and so the calculation process no longer contributes those errors to the estimates. But even though a program no longer assumes that the titration curve of some malt is linear it will still give a bad answer if fed data for that malt which says it is. Thus thee problem becomes one of checking up on malt models.

So let's suppose that the ASBC want's to propose a new MOA (method of analysis) Malt 42A "Malt Titration Parameters". A committee would go off and review the literature and come up with a detailed method including details on how finely the malt sample would be ground, whether they would be dried, how much ground malt would be put into what sort of beaker placed in a bath at what temperature, how much water would be added, how many pH readings would be made and when and so on and so on.

A test malt would then be selected and titrated over and over again in the laboratory. The parameters would be extracted from the titration data and their means, standard deviation and coefficients of variation recorded. Error analyses would be done to propagate the uncertainties in the measured parameters through to a pH estimate based on those parameters. If the resultant uncertainty in the pH estimates were too large the method would be abandoned and another approach sought. If the uncertainties determined were small enough then a collaborative study in which samples of this same malt (or actually several malts, more likely) would be sent to several laboratories and they would measure the malt and report their findings back to the committee. If the interlaboratory variations were too large the method would be dropped but were they sufficiently small it would be accepted.

As ASBC has done this for lots of MOAs there are standard statistical procedures for carrying all this out and it is probably clear that a lot of work by a lot of people is required.

 

[Biscuits]

Hmm not trying to beat a dead horse here but I'm still not understanding. We can prove how accurate the software is by using at and taking the actual ph readings to compare. I do it every brew . Unless my meter is incorrect and I calibrate it before use what are we trying to figure out? In 55+ batches brunwater has never been off by .4. is the question specifically directed at people without meters trying to verify if it's correct?

I think the OP is suggesting that most people are making unintentional errors along the way and giving false validation to existing mash pH software. So, when you take your readings to validate your own water calculations, you might be using your meter incorrectly, or it isn't calibrated correctly, or any other number of variables that can go wrong, which ultimately results in your belief that your chosen calculator is correct when it is not. I dunno, I am just gonna default to what AJ and Martin say, drink my beer, and shut up.

 

[Blazinlow86]

I think the OP is suggesting that most people are making unintentional errors along the way and giving false validation to existing mash pH software. So, when you take your readings to validate your own water calculations, you might be using your meter incorrectly, or it isn't calibrated correctly, or any other number of variables that can go wrong, which ultimately results in your belief that your chosen calculator is correct when it is not. I dunno, I am just gonna default to what AJ and Martin say, drink my beer, and shut up.

Fair enough. I can confidently say the software I use is within .04 of the target and is verified each time using a decent (100$) meter and it's calibrated at each use. I've taken enough samples to feel it's accurate for tg homebrew scale. I guess the only way to know better if you were super obsessed with it would be some sort of lab. I guess for argument's sake one could say at the end of the day the only thing that really matters is when you actually measure your pH whether you're in the range or not. Even if you're making entry errors into your software and you still come out where you want to be it doesn't really make much of a difference. Cheers

 

[ajdelange]

At the ±0.4 level why bother? Throw the meter and software away and make your brew day easier.

 

[Blazinlow86]

At the ±0.4 level why bother? Throw the meter and software away and make your brew day easier.

Lol good point .04

 

[Silver_Is_Money]

I think the OP is suggesting that most people are making unintentional errors along the way and giving false validation to existing mash pH software. So, when you take your readings to validate your own water calculations, you might be using your meter incorrectly, or it isn't calibrated correctly, or any other number of variables that can go wrong, which ultimately results in your belief that your chosen calculator is correct when it is not. I dunno, I am just gonna default to what AJ and Martin say, drink my beer, and shut up.

This goes hand in hand with my contention that if you are sampling before the 30 minute mark (at the earliest), you are getting a false reading. And since most recipes are for beers which reside on the light to low-mid darkness side of the SRM color scale fence, most of such false pH readings will be on the markedly low side, as for these beers the pH rises during the mash. The forum member who recently watched (and charted) his Blonde Ale mash as it progressed from pH 5.42 at the 15 minute sampling, to 5.57 at the end of the mash, to pH 5.63 by the end of lautering proved this. If he was shooting for 5.4 and measured 5.42 at the 15 minute mark he would have vindicated his software (and most of us would simply stop measuring at this juncture, being happy as larks that the software was correct). But in reality the software would in this case be way off, with a fair amount of additional acid being actually required, such that the mash stage terminate at a pH of 5.4. And if you do not do a mash out step to put an end to it, some minor level of saccharification will continue through lautering.

 

[Silver_Is_Money]

To establish whether or not there actually can ever be any hope for statistical relevance in regard to any of the extant gen 1 mash pH assistant software solutions, I place a call to A.J. deLange for his gen 2 software solution to this, a hypothetical SMaSH:

A.J., given this hypothetical and simple SMaSH recipe case, how much 88% lactic acid is actually required to achieve an honest mash pH of 5.4 on a proton basis?

Case:
5 Kg. (11 lbs.) of Maris Otter, measured to have a DI_pH of 5.73 (as was recently done by member 'cire')
17 liters (4.5 gal.) of RO mash water with 50 ppm Ca++, zero Mg++ ppm, and zero alkalinity ppm

This will establish the gen 1 vs. gen 2 software baseline for validity only for the range of light colored brews. Similar standardized example/baseline test cases must then be established for (at a minimum) a mid-SRM recipe (say 15 SRM), a dark recipe (say 30 SRM), and a very dark recipe (say 60 SRM) standardized grist batches. Then all extant gen 1 software can be also compared with hard science for standardized test recipes spanning a range or relevant SRM colors. For the dark brews a standardized percentage of caramel/crystal and roast malts must also be established, as well as a standardized grist. For all of the "standardized" test cases I propose that the same grist weight and mash water conditions (Ca++ content and volume) should be retained as for this light brew test case. And I also propose that the very same 5.73 DI_pH Maris Otter base malt be the foundation for all of the standard grist by SRM color cases.

Then lastly and down the road a bit 70 ppm of alkalinity should be applied to all of these standardized SRM recipes, to test all of gen 1 softwares ability to properly handle added alkalinity.

 

[ScrewyBrewer]

This goes hand in hand with my contention that if you are sampling before the 30-minute mark (at the earliest), you are getting a false reading.

Silver_Is_Money just a thought. But over a year ago after reading AJ's paper it was determined the benchmark for drawing cooled mash pH samples was at the 20 minute mark. Why the sudden shift who determined 30 minutes and how was this communicated to anyone?