Why assigning "HARD" values to "ALL" grains and adjuncts is merely blowing smoke

[Silver_Is_Money]

Meaning here hard values for pHDI and BC, leading (via computation) to hard mEq/Kg_pH acidity valuations.

1) Mainly because "HARD" data for 99+% of these "specific" grains and flakes, and adjuncts, etc... simply and factually doesn't exist.
2) What scant data does exist, when plotted, looks much like a scatter chart, with "inherently" low valuation as to statistical confidence.
3) Even if a pHDI is given by the maltster, it is 100% absolutely useless unless the BC (buffering capacity) is also given, and "specific" BC's are never provided by the maltsters.
4) Annual, seasonal and regional differences within even the very same grain varietal mean that "HARD" valuations are an utter impossibility.
5) Those providing the "HARD" data don't even come close to each other as to their derived critical values when looking at the very same brand and type of malt, grain, or adjunct.

What has been laughed at recently is the use of color whereby to assign the critical values (nominally, via the plotted values 'data regression') is in fact all that one can really go by. When the seemingly as if scatter charts are evaluated mathematically, plotted, and then statistically regressed to 'nominal' math model (equation or formula) values, (as opposed to "HARD") values, these charted values have (surprise) "color" as one chart axis. So to laugh at color as a participant in pHDI, BC, and mEQ/Kg_pH is 'ahem' laughable.

Those who are laughing propose to provide (as a superior method to that of merely deriving 'nominals' from color) highly "specific" (or "HARD") values for pHDI, BC, and mEq/Kg_pH acidity for every last brand name and malt or grain or adjunct type on this planet. One might ask: If such data for the most part simply doesn't exist, and what data does exist looks like a scatter chart with low math model (or curve matching equation) statistical confidence, then how is this possible?

Answer, it isn't! And thus such programs that offer specific item selections right down to malt, grain, etc's brand name for the many hundreds of beer making ingredients are merely blowing smoke. And they are blowing it at you. And for a reason. The real intent of such software is to sell the sizzle as opposed to the steak (or meat) of the programming, and to dazzle the unaware end user.

You are no longer unaware.

 

[Deadalus]

...
2) What scant data does exist, when plotted, looks much like a scatter chart, with "inherently" low valuation as to statistical confidence.


What has been laughed at recently is the use of color whereby to assign the critical values (nominally, via the plotted values 'data regression') is in fact all that one can really go by. When the seemingly as if scatter charts are evaluated mathematically, plotted, and then statistically regressed to 'nominal' math model (equation or formula) values, (as opposed to "HARD") values, these charted values have (surprise) "color" as one chart axis. So to laugh at color as a participant in pHDI, BC, and mEQ/Kg_pH is 'ahem' laughable....

Thank you for making me aware as I was unaware of this discussion.

It would be great if you could provide some examples of this data and where the regressions performed poorly and better for the methods you have described.

 

[McMullan]

Meaning here hard values for pHDI and BC, leading (via computation) to hard mEq/Kg_pH acidity valuations.

1) Mainly because "HARD" data for 99+% of these "specific" grains and flakes, and adjuncts, etc... simply and factually doesn't exist.
2) What scant data does exist, when plotted, looks much like a scatter chart, with "inherently" low valuation as to statistical confidence.
3) Even if a pHDI is given by the maltster, it is 100% absolutely useless unless the BC (buffering capacity) is also given, and "specific" BC's are never provided by the maltsters.
4) Annual, seasonal and regional differences within even the very same grain varietal mean that "HARD" valuations are an utter impossibility.
5) Those providing the "HARD" data don't even come close to each other as to their derived critical values when looking at the very same brand and type of malt, grain, or adjunct.

What has been laughed at recently is the use of color whereby to assign the critical values (nominally, via the plotted values 'data regression') is in fact all that one can really go by. When the seemingly as if scatter charts are evaluated mathematically, plotted, and then statistically regressed to 'nominal' math model (equation or formula) values, (as opposed to "HARD") values, these charted values have (surprise) "color" as one chart axis. So to laugh at color as a participant in pHDI, BC, and mEQ/Kg_pH is 'ahem' laughable.

Those who are laughing propose to provide (as a superior method to that of merely deriving 'nominals' from color) highly "specific" (or "HARD") values for pHDI, BC, and mEq/Kg_pH acidity for every last brand name and malt or grain or adjunct type on this planet. One might ask: If such data for the most part simply doesn't exist, and what data does exist looks like a scatter chart with low math model (or curve matching equation) statistical confidence, then how is this possible?

Answer, it isn't! And thus such programs that offer specific item selections right down to malt, grain, etc's brand name for the many hundreds of beer making ingredients are merely blowing smoke. And they are blowing it at you. And for a reason. The real intent of such software is to sell the sizzle as opposed to the steak (or meat) of the programming, and to dazzle the unaware end user.

You are no longer unaware.

But the range in variation of a given ingredient in space and time doesn't vary that much that even an assumed average for a given type of grain isn't of some value, right? Biology usually has a bigger effect than environmental heterogeneity. I'm not saying developers of 'fancy' software know this.

 

[Silver_Is_Money]

Thank you for making me aware as I was unaware of this discussion.

It would be great if you could provide some examples of this data and where the regressions performed poorly and better for the methods you have described.

One of the best accumulations of the available data can be found within D. Mark Riffe's published paper titled "A Homebrewing Perspective on Mash pH III: Distilled-Water pH and Buffering Capacity of the Grist". From Riffe's seminal work (which garners much praise from me) one will soon see that there isn't much data at all. Riffe's work can be freely downloaded from his website. I have much of this same data as given to me upon request by Briess, and you are certainly capable of requesting the same of them, but mine was not provided to me with an OK to simply disseminate it publicly.

There are no publicly available regressions of this data that I'm aware of. Everyone has to plot it via applying (generally Lovibond or EBC) color on one axis, and either pHDI or BC on the other axis, and then regress it themselves, both for pHDI and BC. The problem is that (as you will see when you are given such maltster data, and/or look at all of the scant data that Riffe was able to collect) some to much of the "HARD" data simply doesn't seem to "fit" the mainstream trend at all, and this brings me to a few major points which I forgot to include above.

Therefore:

6) Everyone must, via their own inherent 'confirmation bias(es)', cherry pick as their bias leads whereby to include and/or exclude the factual "HARD" data, in a rather false, yet seeming logical fashion whereby to be left with "better fitting" data across the color spectrum whereby to regress it. The "HARD" data (which was scant to begin with, and is now really scant) then can be regressed via a multitude of means (regression math models), and it is from this 'regression method pool' whereby ones regression model of choice (often via yet more 'confirmation bias', and/or exhaustion, or laziness) must be selected. The best regression model derives the best confidence level of having achieved a good data fit to the model, but does everyone go through the regression options whereby to discover the best option?

7) I've chosen (for my own software) what I've referred to as a (sort of stellar like) "Main Sequence" whereby to regress such (primarily) barley malts into a single 'main sequence fitting' equation via their 'nominal' color (either as Lovibond or EBC). In my software these "Main Sequence" malts/grains are described as simply "Not Categorized" within the category selection "Drop-Down" column. This means of fit (I.E., Not Categorized) takes care of a vast majority of malts. For a group of malts classified as "Caramel/Crystal" a sub-Main Sequence branch evolves when plotted and lends itself to it's own color based regression equation, which is activated upon selecting "Caramel Crystal" via the Drop-Down. For other ingredients different "Drop-Downs" are to be selected. But in general, if no drop-down presents itself then "Not Categorized" is to be selected. The classification field (cell) remains blank for malts that are not categorized. Other category selections appear visibly in the field. This methodology appears unique (to my knowledge) to my software. I care little as to how others handle this, but it particularly irks me when software presents a facade of knowing far more than it is possible to know.

8) It does appear to me that at least some if not much of the well known mash pH assistant software is applying very simplistic linear modeling, as opposed to non-linear regression modeling. The data itself is clearly not linear, and this seems to lead to either guessing or linear regressing, rather than non-linear regressing, with yet more confirmation bias. And it means that the main sequence must be broken down into additional smaller sequence categories whereby to 'seemingly' (as opposed to factually) fix the data that makes no sense when guessed at, or even regressed linearly, across the entirety of the Lovibond or EBC color spectrum.

9) There is yet another major form of confirmation bias, and that is the bias of the end users. Word of mouth lends support to such confirmation bias, such that (right or wrong) a sort of support group mentality replaces science, and everyone who "Wants to belong" to the perceived 'majority' picks the software that is most hyped by others who need to belong within a comfort zone of peer (pressure?) origin. After all, if everyone says software 'X' is best, then it must be so, right? It's sort of like an old saying that a billion screaming Chinese can't be wrong, can they? And worse, this support group often denigrates the work of others in a fashion that goes something like this: Popularity contest winner 'Software X' predicts 'Y', and other available software does not seem to be in agreement with 'Y', so it inherently must hands down be faulty software.

"Science via majority consensus" is historically extremely misleading, to down right dangerous, or even on occasion life threatening. What might have happened only a few scant centuries ago if you went around saying that the earth is round, or if you suggested that the earth revolves around the sun, or that the earth is not the center of the universe. Going against the majority can be life threatening.

 

[McMullan]

Sounds more like pseudo science than science.

 

[Silver_Is_Money]

Sounds more like pseudo science than science.

I've heard that nearly 80% of sophisticated medical and medicinal study conclusions (leading to a high statistical probability of correctness) can not be independently duplicated and verified to within meaningful statistical confidence. Does that make such medical research (with trials often spanning years) pseudo-science?

Triangle-Tests seem to often lead to conclusions that counter the perceived group mentality 'comfort or support group' expectations. Does that make triangle tests merely pseudo-science? In selecting your answer you are free to hearken back to the paragraph immediately above if/as you choose.

 

[Silver_Is_Money]

But the range in variation of a given ingredient in space and time doesn't vary that much that even an assumed average for a given type of grain isn't of some value, right? Biology usually has a bigger effect than environmental heterogeneity. I'm not saying developers of 'fancy' software know this.

Study the extant data, then attempt to 'validly' make these points from a perspective of knowledge.

That said, I'd "guess" that a majority of those reading this thread will dismiss it carte blanche and return to their peer supported comfort zone of "understanding".

 

[Silver_Is_Money]

In fairness, does my own software have internal equations derived (or perhaps better, tweaked) from personal/inherent confirmation bias and/or data cherry picking?

Yes! It does.

But then again, it rather uniquely provides many end user data/math model "customization" and direct data "override" options whereby to cherry-pick your own biases through which to supplant, improve upon, or otherwise outright replace my biased choices with your own.

 

[Silver_Is_Money]

An example of the poorness of the "HARD" data:

Researcher #1 concludes via testing that the pHDI for Briess 2-Row Brewers is 6.00, and it's BC is 57.2.
Researcher #2 concludes via testing that the pHDI for Briess 2-Row Brewers is 5.55, and it's BC is 46.2.

And another example:

Researcher #1 concludes via testing that the pHDI for Weyermann Munich I is 5.57, and it's BC is 45.6.
Researcher #2 concludes via testing that the pHDI for Weyermann Munich I is 5.44, and it's BC is 52.3.

And on it goes.... Wash, rinse, repeat. But keep in mind that the "HARD" data itself is sparse, and does not exist for more than a small group of such malts, grains, adjuncts. To state that there is insufficient data across the spectrum of such ingredients is an understatement.

Add to that the knowledge that a malts BC is uniquely specific to a single pH targeted titration endpoint, and that the extant "HARD" data often provides zero specific reference as to the targeted ending pH, and "Houston, we have a problem".

Then add to that the knowledge that some BC's are determined via merely adding a fixed titration amount of acid or base (of known strength) to a certain weight of sample, regardless of what the sample may be, and letting the resultant pH endpoint consequently fall randomly where it may, and then via the false presumption of assured titration linearity, presume the BC derived thereby to have a semblance of validity, and you multiply the problem substantially.

 

[Deadalus]

I have the study results from D. Mark Riffe and Mick Spencer downloaded. I am not seeing any indication this was published to a journal. Was this study published just to the web?

Triangle tests derived from 1 or 2 batches of beer are repeated measures on one or two batches of beer. One or two samples is not a sufficient sample size making any such studies "psuedo-science".

What scale of measurement (nominal, ordinal, interval, or ratio) is Lovibond considered? It's at least ordinal but the descriptions I am seeing indicate reference plates were used to determine the value.

If simple linear regression is being used to fit log transformed data than the data are nonlinear to begin with. (If they determine that the slope is significantly different than 0 that is.)

So is there a data set available to illustrate this controversy? I can fit whatever type of regression someone has tried. Or paste in a specific figure where you think somebody has fit the data wrong.

 

[Silver_Is_Money]

These are just a few of the potential reasons why one software may suggest one quantity of required pH adjustment addition, and another may suggest a differing quantity... And for a differing recipe and/or process this may reverse itself.

And why one software may be more correct on occasion (specific to a single set of recipe and process and mineralization impact, and volumes, etc...), and another may be more correct on different of such occasions. Much comes down to a sort of 'Fuzzy Logic' effect, where all of the multiple potentials for error magically cancel (via pure luck) for one, while summing to increase the error for another.

 

[Deadalus]

An example of the poorness of the "HARD" data:

Researcher #1 concludes via testing that the pHDI for Briess 2-Row Brewers is 6.00, and it's BC is 57.2.
Researcher #2 concludes via testing that the pHDI for Briess 2-Row Brewers is 5.55, and it's BC is 46.2.

And another example:

Researcher #1 concludes via testing that the pHDI for Weyermann Munich I is 5.57, and it's BC is 45.6.
Researcher #2 concludes via testing that the pHDI for Weyermann Munich I is 5.44, and it's BC is 52.3.
...

These examples don't illustrate anything. If for instance, the pHDI for Briess 2-row is highly variable in the population, they might be statistically the same. I don't see the poorness there. It would be lousy data if they only took one sample instead of multiple samples. A rule of thumb is 30 samples but it's really a matter of running a power analysis.

 

[Silver_Is_Money]

19th century Philosopher 'Arthur Schopenhauer' once proclaimed that:

All truth passes through three stages. First, it is ridiculed. Second, it is violently opposed. Third, it is accepted as being self-evident.

 

[Deadalus]

 

[Silver_Is_Money]

Have at it. Do the pHDI tests and titrations from which to derive BC's for hundreds of diverse malts, grains, adjuncts, etc... and repeat the tests across a multiplicity of random samples for each until you have gained statistical confidence for all of them, then categorize and then 'best fit' math model them. I would proudly (if still alive) be the first to commend you if you can improve upon the extant situation brought about by scant data.

As for me, I'm an ever aging retiree with several major medical problems. And all I have to work with is the rather scant data that I can find within the public domain plus some scant data provided upon request by Briess.

 

[tracer bullet]

Not so long ago I set out to improve the spreadsheet I use (a modified version of EZ). I gave myself many more grain options, and the pHDI values I could find. I thought it'd be great, then realized how varied those #'s are. I realized that a particular malt will have very different #'s reported - for example by the maltster, AJD's earlier work when he was active, and I believe Riffe as well (I recall reading the white papers but not if I grabbed any numbers from them). I realized right then that it would be futile to have a hard number, and I know as an engineer that even if you make the best measurement in the world, what you are measuring is going to change and so you're never finished.

For what it's worth, I agree, there are no hard numbers. Different sources give different results and I suspect if I made the measurements I'd get yet another result as well. I've become accustomed to comparing my measured pH to the calculated pH, and simply fudging it the next time (say 5.4 is calculated and I get 5.3 in the real world. I know I'm 0.1 off, and so for the next time I brew that recipe or something very similar, I'll change accordingly and take care of it on my own ever after (a little more acid, a little more baking soda, whatever).

I'm interested in making a few pHDI measurements of my commonly used base malts but haven't done so yet. I would hope that could get me closer on the first try.

 

[Silver_Is_Money]

Remember that pHDI's devoid of BC's are useless to the prediction of Mash pH shift. The pertinent equation being:

Delta_pH_Mash = mEq's (of acid or base)/(aggregate_grist_BC x aggregate_KG_malts)

Whereby, for a target mash set at 5.40 pH this becomes:

(pHDI_aggregate_grist - 5.40_target) = mEq's (of acid or base)/(aggregate_grist_BC x aggregate_KG_malts)

Where the computed BC must be calculated at the desired mash pH.

 

[McMullan]

I've heard that nearly 80% of sophisticated medical and medicinal study conclusions (leading to a high statistical probability of correctness) can not be independently duplicated and verified to within meaningful statistical confidence. Does that make such medical research (with trials often spanning years) pseudo-science?

Triangle-Tests seem to often lead to conclusions that counter the perceived group mentality 'comfort or support group' expectations. Does that make triangle tests merely pseudo-science? In selecting your answer you are free to hearken back to the paragraph immediately above if/as you choose.

I wouldn't say failure to replicate results indicates pseudo science. There are a number of good reasons why clinical results might not be confirmed, including comparing apples with pears. Medicine isn't really a science, for ethical reasons. Fortunately. I'd say most clinical studies are flawed by awful biases. But value can sometimes be teased out by meta-analysis, when enough studies have been done. Then better prediction becomes possible. Pseudo science is different. It's flawed from the outset, consciously or not, and never gets better.

 

[Deadalus]

2) What scant data does exist, when plotted, looks much like a scatter chart, with "inherently" low valuation as to statistical confidence.

You are telling us there is a poor interpretation of data but can't produce any of that data or point to a figure and tell us why it was done wrong. Plus you are saying something looks like a "scatter chart" but just because data is presented in a scatter plot, it doesn't mean it's not useful or valid or whatever you think is poor about it. There are ways to demonstrate that a linear regression is not suited for the data but you aren't really hitting on those.

I also asked you what scale of measurement applies to Lovibond but you didn't say. It's important to know so that the appropriate analysis method is selected.

Have at it. Do the pHDI tests and titrations from which to derive BC's for hundreds of diverse malts, grains, adjuncts, etc... and repeat the tests across a multiplicity of random samples for each until you have gained statistical confidence for all of them, then categorize and then 'best fit' math model them. I would proudly (if still alive) be the first to commend you if you can improve upon the extant situation brought about by scant data.

As for me, I'm an ever aging retiree with several major medical problems. And all I have to work with is the rather scant data that I can find within the public domain plus some scant data provided upon request by Briess.

Since it is pretty nonexistent data I don't doubt I could improve upon it and it wouldn't necessarily follow the sequence you are describing. First off, rather than reporting a single measured point, one can generate a sample mean and sample variance and standard error for the mean from a small sample. That would be better than just one point. Then one would have some idea about the variability of what's being measured. You could do some pilot studies to see if certain grains are or aren't different and perhaps design effective experiments that do describe classes of grains well (or not).

You are assuming that everything is different but don't can't/won't show the data and accusing other of being incorrect for making assumptions without pointing to where they have done so erroneously. I just think if you could show the data, any data, we could talk about it.

 

[Silver_Is_Money]

I wouldn't say failure to replicate results indicates pseudo science. There are a number of good reasons why clinical results might not be confirmed, including comparing apples with pears. Medicine isn't really a science, for ethical reasons. Fortunately. I'd say most clinical studies are flawed by awful biases. But value can sometimes be teased out by meta-analysis, when enough studies have been done. Then better prediction becomes possible. Pseudo science is different. It's flawed from the outset, consciously or not, and never gets better.

Then why did you infer that Mash pH prediction via math modeling is merely pseudo-science?

 

[Silver_Is_Money]

You are telling us there is a poor interpretation of data but can't produce any of that data or point to a figure and tell us why it was done wrong. Plus you are saying something looks like a "scatter chart" but just because data is presented in a scatter plot, it doesn't mean it's not useful or valid or whatever you think is poor about it. There are ways to demonstrate that a linear regression is not suited for the data but you aren't really hitting on those.

I also asked you what scale of measurement applies to Lovibond but you didn't say. It's important to know so that the appropriate analysis method is selected.


Since it is pretty nonexistent data I don't doubt I could improve upon it and it wouldn't necessarily follow the sequence you are describing. First off, rather than reporting a single measured point, one can generate a sample mean and sample variance and standard error for the mean from a small sample. That would be better than just one point. Then one would have some idea about the variability of what's being measured. You could do some pilot studies to see if certain grains are or aren't different and perhaps design effective experiments that do describe classes of grains well (or not).

You are assuming that everything is different but don't can't/won't show the data and accusing other of being incorrect for making assumptions without pointing to where they have done so erroneously. I just think if you could show the data, any data, we could talk about it.

It seems as as if you have read essentially none of what I have posted. I've pointed you to much of the extant data. And I've expressed that you must plot it yourself whereby to see with your own eyes the degree of data point scatter. And I've expressed that I do not use simple linear regression, and that pHDI and BC lead to malt acidity which is scaled vs. color. And I've told you that malt acidity (as derived from BC and pHDI) is also scaled to pHDI and a target mash pH. And then I gave you my core pH shift computing equation (which plainly reveals this integrated relationship) as a starting point. And then I stated that the math modeling of BC's can be derived (albeit to a rather low confidence degree, with this being a reason why softwares deviate in prediction) via the non-linear regression scaling of malt acidities to their respective malt colors. And then I speculated that 'others' use simple linear modeling.

 

[McMullan]

19th century Philosopher 'Arthur Schopenhauer' once proclaimed that:

That sounds about right, psychologically, but 'truth', scientifically speaking, is a probability, often mistaken for a 'fact'. In a binary world, people are blind to probabilities and celebrate beliefs they've successfully rejected mainstream 'facts'. Reinforced by herd mentality and belief in more attractive alternative realities. Usually the aim of psuedo scientists, trying to sell us something. As long as it's not the truth.

 

[Silver_Is_Money]

Ah, we have 'perhaps' at last found common ground. Mash pH prediction is predicated upon a multitude of probabilities. I hope I've made that quite clear. Sometimes the probability stars align, and sometimes they don't. And you can't use a small set of aligned probabilities whereby to state that for all possible grists, procedures, weights, volumes, etc... one software is de-Facto superior to another for all possible cases. One must look elsewhere whereby to speculate as to superiority. I suggest core math modeling validity as at least a start. But even if the core math modeling is sound, the variables within must be sound. And they are not, because the extant data is not sound.

 

[McMullan]

Then why did you infer that Mash pH prediction via math modeling is merely pseudo-science?

I was referring to

"Science via majority consensus"

Where the consensus didn't seem to be a scientific consensus. I don"t consider maths to be a science regardless. So I would call it a pseudo science.

 

[Silver_Is_Money]

I was refuting science via consensus.

 

[Silver_Is_Money]

In all of this the core reason for this thread has been to say that those softwares that present the outward facade of knowing the core BC, pHDI, and thereby malt acidity values for each and all of the myriads of hundreds of beer ingredients subject to mash pH alteration are merely blowing smoke. The smoke that wafts from the sizzle of the software as it dazzles the human senses via its apparent beauty.

 

[Deadalus]

It seems as as if you have read essentially none of what I have posted. I've pointed you to much of the extant data. And I've expressed that I do not use simple linear regression. And I've told you that malt acidity (as derived from BC and pHDI) is scaled to pHDI and a target mash pH. And then I gave you my core pH shift computing equation as a starting point.

You keep dancing around things and not answering questions.
Lovibond what measurement scale is it?

You pointed me to a report which I asked if it was actually published to a journal or not you ignored. I see now though, you were actually pulling the data from there in the Appendix. Ok we're getting some where. In the Appendix, Table V, the Briess 2-row data (Pils/Lager) you mentioned are from a study by Bies and a study by Guerts. The quality of those values could be better determined by reading those studies. (We could have determined something about them if for each line there was a sample size for each and a standard error. Unless of course the sample size is 1.) Do you find that the values presented for that category for pHi (5.76+-0.18) and -Bi(45.5 +/-8.3) are poor?

 

[Silver_Is_Money]

You keep dancing around things and not answering questions.
Lovibond what measurement scale is it?

You pointed me to a report which I asked if it was actually published to a journal or not you ignored. I see now though, you were actually pulling the data from there in the Appendix. Ok we're getting some where. In the Appendix, Table V, the Briess 2-row data (Pils/Lager) you mentioned are from a study by Bies and a study by Guerts. The quality of those values could be better determined by reading those studies. (We could have determined something about them if for each line there was a sample size for each and a standard error. Unless of course the sample size is 1.) Do you find that the values presented for that category for pHi (5.76+-0.18) and -Bi(45.5 +/-8.3) are poor?

My god, man! I've given you leagues well more internal insight (in this thread, as well within many more and far more detailed of my threads to be found here at HBT and/or at it's Brewers Friend affiliate) than well most of the rest, who mainly offer you nothing, sans perhaps upgrades to pay versions. This is the age of the internet. Use it if you need to know the history and science behind Lovibond and EBC and SRM color, etc....

 

[Silver_Is_Money]

You can essentially design your own mash pH modeling software package from a collection of my posts. And at least one person privy to inside info at a major forum discussing such matters has (via the use of my posts) abandoned the forums modeling and replaced it with mine for his own use.

My modeling is after all nigh-on fully in the public domain. Who else offers you that?

 

[Silver_Is_Money]

When error and/or improvement is found within my software, I publicly apologize for said error and offer a correction for it with an accompanying explanation. And ditto I try hard to explain the essentials of my intended improvements. And if my intended improvements fail, I also apologize for that and either back them out or correct them. Is anyone else who is playing within this highly nitch software arena willing and open to doing this? Are they willing to openly expose themselves to public (and/or private) ridicule for error such as I am?

 

[Deadalus]

My god, man! I've given you leagues well more internal insight (in this thread, as well within many more and far more detailed of my threads to be found here at HBT and/or at it's Brewers Friend affiliate) than well most of the rest, who mainly offer you nothing, sans perhaps upgrades to pay versions. This is the age of the internet. Use it if you need to know the history and science behind Lovibond and EBC and SRM color, etc....

I want to see whether or not you know how to model the data you are talking about. If you can't answer a simple question about the measurement scale of one of the variables I have to question your statistical acumen. It is up to you, the scientist, to provide enough detail that the reader can determine whether the study has been conducted correctly. You have started this thread calling out other people's methods yourself. I don't get the feeling you have a good understanding of uncertainty in data and ways to address it.

It also seems like you are importing an argument from other forums and you have written a manifesto. I can probably help you to shed light on the issues but you have to start by cooperating and answering questions. You mentioned in #6 about deciding the best analysis method. I always look at data in different ways and run tests and diagnostics to determine whether the models are appropriate for my data and I give a hard look at data that is presented to me as well.

 

[balrog]

what you are measuring is going to change

Truer words were never spoken.

 

[Silver_Is_Money]

I'm not the one who has done the titrations (for BC determination) plus pHDI measurements that lead to abysmally scattered data. Look at the data yourself. How am I expected to know their methodologies whereby to explain why when all of such data is used whereby to compute BC and thereby malt acidity when plotted in accordance with color it gives merely a highly scattered appearance?

 

[Silver_Is_Money]

I want to see whether or not you know how to model the data you are talking about. If you can't answer a simple question about the measurement scale of one of the variables I have to question your statistical acumen. It is up to you, the scientist, to provide enough detail that the reader can determine whether the study has been conducted correctly. You have started this thread calling out other people's methods yourself. I don't get the feeling you have a good understanding of uncertainty in data and ways to address it.

It also seems like you are importing an argument from other forums and you have written a manifesto. I can probably help you to shed light on the issues but you have to start by cooperating and answering questions. You mentioned in #6 about deciding the best analysis method. I always look at data in different ways and run tests and diagnostics to determine whether the models are appropriate for my data and I give a hard look at data that is presented to me as well.

I've given you the core essence of the publicly available data, and also the means whereby to derive BC and from that malt acidity. It's up to you to regress it (and to pick and choose your own biases) whereby to determine the best data and the best fit regression method. I've further told you (and all) that data provided to me by Briess upon personal request is not privy to the public domain. So I couldn't present you with a full (albeit still scant and scattered) data picture even if I wanted to.

 

[Deadalus]

I'm not the one who has done the titrations (for BC determination) plus pHDI measurements that lead to abysmally scattered data. Look at the data yourself. How am I expected to know their methodologies whereby to explain why when all of such data is plotted in accordance with color it gives merely a highly scattered appearance?

You are qualifying "scattered data" with abysmally. Produce an actual plot here where we can see it that leads you to use that descriptor.

 

[Silver_Is_Money]

You are qualifying "scattered data" with abysmally. Produce an actual plot here where we can see it that leads you to use that descriptor.

Plot it yourself, sans for those malts/grains that fail to align with the "main sequence", such as for Caramel/Crystal, and the scatter will become visually self evident. Then regress it into an equation for BC plus an equation for pHDI (both with respect to color) and the 'abysmal' statistical confirmation (R and R^2) will confirm the scatter.

We are talking in circles. Further of such circular discussion will not garner a response. I'm not seeking your data regression assistance. And I have no obligation to submit to demands merely at your asking. If I made such demands of you out of thin air, would you openly comply as if owing me such information (which in my case contains embedded within confidential data mixed with public domain data) to me?

 

[Deadalus]

Plot it yourself, and the scatter will become self evident. Then regress it into an equation for BC plus an equation for pHDI (both with respect to color) and the 'abysmal' statistical confirmation (R and R^2) will confirm the scatter.

We are talking in circles. Further of such circular discussion will not garner a response.

It is absolutely not up to me to do your work for you. If you can't support it with specific examples you shouldn't be saying it. If you don't know the measurement scale of the data you are using you shouldn't be modeling it.

Also, Riffe and Spencer do include the citations for the Appendix data. They are not Researcher #1 and Researcher #2. If you are going to critique those values, you should do your homework and tell us why their values aren't appropriately determined. It's entirely possible that if tested they are statistically the same. It could also be that they are statistically different but the difference could be unimportant. You know what else, they really shouldn't be exactly the same because unless the population variance is really tight, one should expect different sample means. And if each were taking just one sample, they really ought to be different for continuous data.

I asked you to point to one plot that was poor in your opinion and you tell me I have to plot it out myself. Not a good way to make a scientific argument. "Well I can only prove my point if you do the experiment."

 

[Silver_Is_Money]

When and where did I suggest that Riffe and Spencer were what I referred to as the researcher #1 and researcher #2 sources for the public domain data, some of which precedes them by years. Briess provided me with some of this data before I even knew who Riffe and Spencer were. I have no idea as to whether they received the Briess contribution to their data before or after I did. Much credit for my core modeling methodology goes back to AJ deLange. Please do not accusationally impose such falsehood upon me.

My desire is for a moderator to review this accusation.

 

[Yooper]

I've been following along. We can either continue the discussion if all parties can agree to be less antagonistic, or close the whole thing.

 

[Silver_Is_Money]

The best fit(s) to the data are logarithmic, and thus the best regression methodology whereby to assign a 'nominal' pHDI and also a 'nominal' BC based upon malt (and for some cases grain) color is to apply a logarithmic regression. I am not the first to have recognized this, so I take no credit.