Validating Residual Correction in Fresh-Food Companion-Animal Diets: A Paired-Arm Reliability Ablation, a Stochastic Reliability Framework, and a Tissue-Kinetic Safety Model

Manuscript 8 of 8 in the BARC Protocol specification series · Release B-2.1.0 · Version 3.0 · September 2026

Author. Prasanna M, Growlrr Foods Pvt Ltd, Chennai, India. Correspondence: support@growlrr.com. ORCID: 0009-0003-6812-347X.

DOI. 10.5281/zenodo.22647623 — the concept DOI, which always resolves to the current version of this record · Published. 8 September 2026 · Licence. CC BY-SA 4.0 · Status. Not peer-reviewed as of the date of publication, 8 September 2026.

Competing interests and funding. The author is the founder of Growlrr Foods Pvt Ltd, a pioneer and proponent of the BARC Protocol for pet foods. BARC stands for Biologically Appropriate, Residually Corrected, and is conceptualised as an open protocol under a Creative Commons Attribution–ShareAlike licence, for like-minded collaborators around the world to formulate and develop. Growlrr Foods Pvt Ltd develops and commercialises a BARC-based residual-correction product, BowlBalancer. The work was funded entirely by the company; no external funding was received. No independent laboratory, institution or reviewer has verified the results reported here. Readers should weigh these facts alongside the analysis, which is published in full — including its limitations and open items — precisely so that it can be checked.

Framework. This work applies BARC — Biologically Appropriate, Residually Corrected — an open protocol for correcting a biologically appropriate fresh base to a published nutrient standard, released under a Creative Commons Attribution–ShareAlike licence (CC BY-SA 4.0) at growlrr.com/barc-protocol (concept DOI 10.5281/zenodo.22648597) so that others may formulate and develop against it. BARC is the protocol; the correction specified and validated here is one implementation of it.

Terms. A glossary of the statistical and nutrition vocabulary used across this series — requirement, floor, safe upper limit, Wilson bound, N-limited, draw, stress multiplier, retention factor, trituration and the rest — is in the series cover note (10.5281/zenodo.22647606), section 4.1.

Computational tooling and AI assistance. The model is implemented in TypeScript under Node.js with no external numerical dependencies. Claude (Anthropic) was used as a coding and analysis assistant in developing this work and preparing these records; it was not used to generate results, all of which were produced by executing the model. See the series cover note (10.5281/zenodo.22647606), section 6.1.

Plain-language material. A non-technical account of the validation programme is at growlrr.com/how-we-validated-homemade-pet-food; what "complete and balanced" means against a published standard is set out at growlrr.com/pet-nutrition/complete-and-balanced; and the interactive diet checker described in §6 is at growlrr.com/fix-my-pets-diet. The public reference list behind these claims is at growlrr.com/pet-nutrition-references.

Version note. This is version 3.0 of the methodology record and supersedes version 2.0 of 29 August 2026, which was published under CC BY-NC-ND 4.0; those terms continue to apply to that version.

Abstract

Fresh, home-cooked diets for cats and dogs are biologically appealing but, evaluated against an established nutrient standard, are routinely incomplete. A residual-correction approach — a biologically appropriate fresh base plus a calculated micronutrient correction — can close that gap, but the correction must be shown to hold across the real-world variability of home ingredients, over months of feeding rather than on paper for one idealised bowl, and demonstrably because of the correction rather than because of the base.

We describe three complementary methods and report a controlled ablation of the correction itself.

(1) A paired-arm reliability ablation. The identical simulation — same diet, same variability model, same pseudo-random seed, same draws — is executed twice, once with the correction applied and once with its contribution reduced to a nutritionally null value. This isolates the correction as the single manipulated variable. Across 100,000 simulated feeding days per arm, the corrected diet clears every nutrient floor on every draw (zero observed failures; Wilson-95 lower bound 99.996%, a value limited by sample size rather than by the diet), while the same balanced fresh base without correction clears every floor on no draw at all (Wilson-95 = 0%). The uncorrected failure is neither marginal nor stochastic: the base misses 11 to 16 nutrient floors simultaneously, several at a literal zero percent of requirement, so the joint criterion cannot be met in any draw at any sample size.

(2) A stochastic reliability framework. The Monte-Carlo procedure underlying both arms, which resamples realistic ingredient, bioavailability, potency and fill variability and reports the fraction of simulated daily bowls clearing every floor without approaching a ceiling, with a Wilson confidence bound.

(3) A tissue-kinetic safety model. A one-compartment toxicokinetic account of chronic micronutrient accumulation, with homeostatic absorption regulation and first-order clearance parameterised from published literature, showing that a corrected diet remains below safe upper limits over long-run feeding.

Together these model results are consistent with adequacy in a fresh diet being a distributional and time-integrated property that ingredient selection alone does not deliver, and they isolate the correction as the manipulated variable within the simulation. Because both arms draw on the same composition inputs, the ablation demonstrates internal dependence on the correction term; it is not a causal effect measured in animals. Formulation proportions and engine internals are excluded by design.

Keywords: BARC Protocol · residual correction · homemade dog food · homemade cat food · companion animal nutrition · ablation study · controlled comparison · Monte-Carlo simulation · Wilson confidence interval · sample-size limitation · convergence · nutritional adequacy · home-prepared diet · toxicokinetics · safe upper limit · NRC 2006

1. Introduction — background and scope

Home-prepared diets are widely reported to fall short of published nutrient requirements when assessed rigorously [1]; the biological appeal of fresh food does not, by itself, make it complete. A residual-correction diet answers this by pairing a biologically appropriate fresh base with a small, calculated correction that closes the measured gap to a target standard [2,3]. Two validation questions then arise, and neither is answered by a deterministic spreadsheet:

1. Reliability under variability. Home ingredients vary by cut, season, source and cooking loss. Does the corrected diet clear the requirement across that distribution, not merely at nominal values?
2. Long-run safety. Fat-soluble vitamins and certain trace minerals accumulate. Does the diet stay under safe upper limits over months of daily feeding, given physiological buffering?

A third question is prior to both: is the correction what makes the difference? A diet could in principle be adequate because the base was well chosen. That question is answered in §3.

This is a methods record, not a product disclosure. The correction's composition and the software computing it are proprietary and are not included; any reproduction represents the correction abstractly, as a mass that closes the measured gap.

2. The adequacy target and the energy model

2.1 Nutrient target

Adequacy is assessed against the NRC 2006 requirements for adult maintenance, and by extension the derived AAFCO and FEDIAF profiles [2,3,4], species-specific, expressed per unit of metabolisable energy and scaled to the animal's daily energy requirement.

2.2 Allometric energy model

Daily maintenance energy is sized allometrically:

• Dog: MER = 98 · BW^0.64 · M (kcal/day)
• Cat: MER = 90 · BW^0.67 · M (kcal/day)

where BW is body weight in kilograms and M is a dimensionless multiplier for neuter status, season, body condition, housing and activity. The dog exponent 0.64 follows the measured whole-body calorimetry of Burger and Johnson [5] — 22 dogs, seven breeds, 5.8–48.8 kg, 678 · W^0.64 kJ/day resting, r = 0.96 — rather than Kleiber's theoretical interspecific 0.75 [6]. The cat exponent 0.67 follows NRC 2006.

The coefficient cannot be read without its multiplier band:

Quoting a coefficient alone is meaningless, since 90 × 1.1 and 100 × 1.0 differ by 1%.

Food-energy buffer. The fresh base is sized not to maintenance energy but to

targetKcal = 1.10 × MER

a deliberate 10% headroom absorbing day-to-day variation in owner portioning and ingredient energy density. Note that MER already contains M; the buffer multiplies MER, and applying M a second time would double-count it. All reliability figures below are computed on this basis, and the buffer is never used in a floor lookup.

Per-kilogram nutrient targets scale inversely with M: a fixed-dose correction delivers proportionally less per calorie to a more active animal. The reliability analysis accounts for this by sampling M across the band rather than evaluating at the anchor. The full model, its provenance and its external triangulation are Manuscript 4.

2.3 Biological-appropriateness boundaries

Independently of adequacy, the fresh base is required to satisfy three quantitative constraints, by species the BARC Protocol (see Framework, above):

These bound the base. The correction supplies micronutrients — including the calcium a boneless cooked base lacks and the heat-labile vitamins cooking degrades — not macronutrient structure. Internal diagnostic tooling previously used looser working thresholds; those were a calculator's defaults, and the values above are the published boundaries of record.

3. Methods and results 1 — the paired-arm reliability ablation

3.1 Rationale

Methods 2 and 3 establish that a corrected fresh diet is reliable and stays safe. Neither shows that the correction is what makes it so. Descriptive surveys of home recipes cannot separate these: they observe uncorrected bowls failing, but never hold the bowl constant. The ablation removes the ambiguity by treating the correction as the single manipulated variable in an otherwise identical experiment.

3.2 Design

Two arms per diet. Everything is held constant between them — the ingredient list and gram weights, the energy model, the variability and retention distributions, the draw count, and the pseudo-random seed — so the two arms consume the same stream of random numbers and differ only in whether the correction is applied.

• Corrected arm. The correction is applied at its specified dose.
• Uncorrected arm. The identical pipeline runs with the correction's mass scaled to a nutritionally null fraction of a dose. The correction is ablated, not removed from the code path, so no other behaviour changes.

Two diets bracket the realistic range:

• Balanced fresh base — the recommended multi-protein bowl (poultry, red meat, farmed white fish, egg, dairy, vegetable matter). This is the strongest uncorrected case: a competent, varied, properly cooked home diet.
• Mono-protein base — a single-protein (chicken) bowl sized to full food energy. The worst realistic case.

3.3 Outcome measures

Reliability (Wilson-95). The lower bound of a Wilson score interval [11] on the proportion of simulated days on which every floor is cleared — the joint criterion.

Simultaneous floor misses. The count of distinct nutrients delivered below requirement. This is reported in preference to naming a single binding nutrient. The binder — the worst nutrient in a given draw — is a Monte-Carlo diagnostic answering which nutrient decided this draw; it systematically understates a diet that misses a dozen floors at once, and should not be used to characterise a bowl.

3.4 Result

Across 100,000 simulated feeding days per arm (see §3.6 on why this and not 10⁶), on the release-B-2.0.0 diet and correction:

Uncorrected nutrient delivery, mono-protein base, dog, as a percentage of the NRC 2006 requirement, before and after correction:

{{CHART:ablation|The paired-arm ablation, per nutrient. Both arms are identical populations; the correction is removed in one. Values are percentages of requirement — note that EPA and DHA are absent from the uncorrected base entirely, and that the corrected value for that nutrient (1,026%) is clipped by the axis. A fixed-composition unit dosed by count cannot land every nutrient at 100%; the binding constraint is the safe upper limit, not the floor.}}

3.5 Interpretation

The uncorrected result is not a probabilistic near-miss; it is a structural zero. Six nutrients are delivered at a literal 0% of requirement by a mono-protein cooked base — calcium (the base is boneless), thiamine (destroyed by cooking), vitamin E, choline, manganese and long-chain omega-3. A joint criterion requiring all floors cannot be satisfied in any draw, however many are taken. The Wilson bound of 0% therefore reflects a deterministic property of the diet, not sampling variation, and would not improve with larger N.

Equally, the balanced base — a genuinely good, varied, competently cooked diet — still misses 11 (dog) to 14 (cat) floors simultaneously. Ingredient quality and variety do not close the gap; they narrow it. This is the modelled basis of the claim that biological appropriateness and nutritional adequacy are independent axes (§2.3, Manuscript 1); it is arithmetic on tabulated composition values, not an observation.

The corrected arms clear every floor on both diets. Notably the mono-protein base corrected is as reliable as the balanced base corrected — the correction is not topping up an already-near-adequate bowl; it is carrying the diet across the threshold from a starting point of zero.

3.6 On sample size, and why the reliability bound must be read carefully

A reliability figure of this kind invites a misreading that the method itself should foreclose. Sweeping the draw count on the corrected configuration separates two quantities usually reported together:

The physical quantity converges; the bound does not. Worst-nutrient delivery is stable from 5,000 draws — it moves 0.2 percentage points across a fortyfold increase in N. The Wilson bound rises monotonically, because once observed failures reach zero the limit reduces to

LB = n / (n + z²), with z = 1.96

a function of sample size alone. The table above reproduces exactly at that constant: 5,000/(5,000 + 1.96²) = 99.923%.

We therefore report zero observed failures at N = 100,000, and state the bound as N-limited. Claims of the form "99.996% reliable" are avoided, since the trailing digits encode compute budget rather than diet quality.

A limitation. The simulation is seeded deterministically, so the sweep above is a nested sample on a single stream. It establishes convergence but does not provide a between-replicate variance. Independent replicate dispersion became available separately, and is reported in §4.3.

3.7 Reproducibility and evidence

Both arms run through the same top-level entry point, at a fixed seed, on the published allometric model and the NRC 2006 floors. The complete engine return for every arm — reliability statistics, per-nutrient delivery for all 28 floor nutrients, and the run environment — is preserved as machine-readable evidence, so the tables above can be audited without re-execution. The composition of the correction remains proprietary and is represented throughout as an abstract mass.

4. Methods and results 2 — the stochastic reliability framework

4.1 Procedure

For each of N iterations: sample a bowl — draw ingredient quantities and nutrient contents from distributions reflecting real variability, including natural compositional variance, sourcing and seasonal spread, cooking retention (thiamine retention approaching zero under cooking), and bioavailability discounts for minerals; apply the correction sized to the animal's target; evaluate the delivered vector against every NRC 2006 floor and every defined safe upper limit; record pass or fail. Aggregate to a reliability with a Wilson bound.

Five lognormal noise sources are applied per draw: unit fill mass (CV 0.05), purified-active potency (0.05), grocery portion and composition (0.10), biological food-form bioavailability (0.20), and cooking retention (0.20, scaled per nutrient). A parallel arm tracks delivered-versus-required at the tissue level rather than at intake level (§5).

4.2 Gate

C/R 95/99 — the Wilson-95 lower bound of the daily pass proportion must be ≥ 99% of days at ≥ 90% of the NRC requirement. A batch fails if any nutrient reaches ≥ 80% of its safe upper limit.

4.3 Reported outcome, and its exact scope

Draw counts differ by configuration, and the distinction is load-bearing:

The between-run standard deviation is the quantity §3.6 identifies the nested sweep as unable to provide. Independently seeded streams make ten runs ten replicates rather than ten identical repeats, which is what permits a replicate-based dispersion at all.

Full detail of the 1,000,000-draw run, its audit corpus and the family-configuration results is Manuscript 3.

4.4 Reproducibility

The framework is reproducible from the NRC 2006 floors and safe upper limits (public), the allometric model (§2.2, public), the variability and retention distributions (documented qualitatively here), and a fixed seed. The specific correction vector is proprietary and is represented in any reproduction as an abstract correction closing the nominal gap.

5. Methods and results 3 — the tissue-kinetic safety model

5.1 Rationale

Time-integrated adequacy and chronic accumulation are assessed on a separate axis from the daily snapshot, over a 365-day horizon. Fat-soluble vitamins and certain trace minerals are the relevant risks. This model tests the diet on the time axis.

5.2 Approach

Each nutrient is buffered over its storage horizon. A first formulation uses an exponential moving average parameterised by biological half-life; a second replaces it with a one-compartment model carrying explicit homeostatic absorption regulation and first-order clearance, every parameter taken from published literature:

S_{t+1} = S_t + k_out · (netAbs(I_t) − S_t), netAbs(I) = 100 · (I/100)^p, k_out = ln 2 / t½

Setting p = 1 recovers the moving average exactly, so the kinetic model is a strict generalisation: it can add regulation detail but cannot contradict the floor result. Nutrients that cannot be parameterised run at p = 1, taking no regulation credit.

5.3 Reported outcome

The corrected diet reaches a steady-state body burden below the safe upper limit for all accumulating nutrients over a multi-month horizon. Under deliberate copper over-supply, modelled tissue copper compresses toward requirement rather than accumulating toward the ceiling — the cited absorption exponent for copper is p = 0.34 [7] — plateauing near 155% of requirement at three-fold over-supply, roughly six to seven times below the safe upper limit.

Model structure is disclosed; the correction magnitude feeding it is proprietary. Full parameters, results and the species-extrapolation register are Manuscript 6.

6. Results — the supporting ten-run bowl matrix

Before the paired-arm ablation was built, the same question was asked deterministically across a range of diets: five bowls spanning all-meat to all-plant, each evaluated twice — as fed, and with the correction added. Ten runs in total. The bowl is the base case; the correction is the transformation.

Table 6.1 — Ten-run matrix. Unprimed = the bowl as fed; primed (′) = the same bowl with the correction. "Breaches" counts nutrients below requirement.

The correction closed every breach in all five diets — 9 to 21 breaches, zero remaining, every time.

Four patterns hold across every diet in the matrix, and none of them depends on which bowl was chosen.

Thiamine is zero in every uncorrected bowl. Not low — zero. All five are cooked, and thiamine is assigned no retention through cooking [Manuscript 2 §3.1], so the base contributes none of it and the corrected value is identical (220%) in all five. The thiamine column is the heat-retention assumption made visible.

Calcium-to-phosphorus is inverted in every uncorrected bowl — 0.18 to 0.69 against a target near 1.3 — because all five bases are boneless. Correction moves every one into 1.31–1.86. The all-meat bowl is the worst of the five at 0.18, which is the opposite of intuition and the clearest evidence that meat quality does not confer calcium adequacy.

Taurine and long-chain omega-3 track the animal fraction exactly. Both are 0% in every plant-based bowl and highest in the all-meat one. They are animal-tissue nutrients, and no amount of plant protein substitutes for them.

Protein quality does not follow the breach count. The all-meat bowl has the fewest breaches (9) and by far the best amino-acid margin (Met+Cys 291%); the plant bowls sit at 97–100%, on the line. Breach count and protein quality are different axes, which is precisely the point §6.1 develops.

6.1 What this study produced — the two-axis model

The matrix was not designed to test the classification; it broke it. Under a single-scalar tier rule, A′ and C′ both classified as the top tier.

A′ is egg, curd, rice, dal, tofu and vegetables — no meat at all, amino-acid margin exactly at the recommended allowance, food-borne taurine and omega-3 both zero. C′ is a diverse animal bowl with nearly three times the amino-acid margin and both animal-tissue markers well supplied. Awarding those two the same label is not defensible.

The cause was that a single scalar was deciding a label that claimed much more than the scalar tested. The fix used no new criteria — only a separation of the ones already there into two independent axes: whether the bowl is adequate once corrected, and whether the base is biologically appropriate. That separation is the two-axis taxonomy the framework now rests on, and this matrix is where it came from.

6.2 Status of this study

It is deterministic and descriptive. It evaluates point estimates rather than distributions, and although it varies the bowl it does not hold the bowl constant, so it cannot attribute the closing of a breach to the correction rather than to the recipe. Method 1 supersedes it on the causal question for exactly that reason.

It is retained here for three things it does that the ablation does not: it spans a range of diet types rather than two; its independently derived range of 9–21 breaches corroborates the ablation's 11–16 from a different construction; and it is the study from which the classification in §2.3 was derived.

The instrument is public. Readers can run their own bowls against the same model at growlrr.com/fix-my-pets-diet, adjust any ingredient, and toggle the correction to reproduce the paired structure of this matrix on a diet of their choosing. It is a deterministic point-estimate tool and is labelled as such; it is not the stochastic gate of [Manuscript 2].

7. What is disclosed and what is withheld

Disclosed. The nutrient target and its provenance; the allometric energy model in full — coefficients, exponents, per-species multiplier bands and anchors, and the 1.10 food-energy buffer; the biological-appropriateness boundaries; the three validation methods in enough detail to reproduce the procedure; the qualitative structure of the variability, retention and bioavailability inputs; and, for the ablation, the complete per-arm result set across all 28 floor nutrients for both species, both diets and both arms, as machine-readable evidence.

Withheld (proprietary; not part of this deposit and not licensed). The exact correction formulation and composition — the bill of materials — and the internal balancing-engine source. Any reproduction of the methods represents the correction abstractly.

8. Limitations

• No empirical referent. Every figure here is a model output; no animal has been fed and no finished lot assayed. A non-invasive household observation study is planned to begin closing this, and results will be reported when they exist (series cover note, Limitations and future work).

• The Monte-Carlo cannot see a misweigh. Potency and fill are lognormal coefficients of variation, which does not represent a discrete decimal error on a microgram-scale active. That risk is controlled by process — assayed dilute premixes, geometric dilution, per-lot assay [Manuscript 4] — not by this model.
• Deterministic seeding makes results reproducible and the §3.6 sweep non-independent; only the ten-seed run yields replicate dispersion, and only for the adult corrections.
• Bioavailability and retention distributions are literature-derived and documented qualitatively rather than per-nutrient against primary sources.
• Safe upper limits were accessed through secondary summaries and should be verified against the primary NRC 2006 volume.
• The ten-run matrix is deterministic and descriptive and does not support a causal claim (§6.2).
• The energy model is empirically grounded over 5.8–48.8 kg; outside that range it is a principled extrapolation [Manuscript 3].
• Healthy animals only. Not a clinical or therapeutic system, and no disease claim is made.
• Not peer-reviewed as of the date of publication (8 September 2026). Should it later be reviewed or independently replicated, that will be recorded in a subsequent version of this record.

9. Provenance

This work synthesises two prior traditions and names its debts. The principle of biological appropriateness derives from BARF (Biologically Appropriate Raw Food), coined by the veterinarian Ian Billinghurst [8]. The discipline of correcting a home diet to a published nutrient standard derives from home-formulation practice established by board-certified veterinary nutritionists, notably the BalanceIT system of Sean Delaney [9,10]. The present contribution is the two-axis standard, its published quantitative boundaries, and the distributional and time-integrated validation methodology above — with a boneless, cooked base whose micronutrient residual is closed by a validated correction rather than by feeding raw bone.

10. Conclusion

This record states the methodology behind the series and reports the ablation that tests whether the correction is doing the work attributed to it. Two identical simulated populations are run, differing only in whether the correction is present. The uncorrected balanced base — a varied, competently cooked diet — misses 11 to 14 nutrient floors simultaneously depending on species, and the single-protein base misses 14 to 16; the corrected arms clear every floor on every draw.

The ten-bowl matrix extends this across diets a household might plausibly assemble, and supports the paper's central methodological claim: that biological appropriateness and nutritional adequacy behave as independent axes, since bowls can be varied, fresh and appropriate while remaining short on the nutrients a cook cannot see.

What the ablation demonstrates is internal dependence on the correction term within the simulation. Because both arms draw on the same composition tables, it is not a causal effect measured in animals, and the deterministic seeding of the convergence sweep makes that sweep a nested sample rather than independent replication. Those limits are stated because the alternative — presenting arithmetic on tabulated values as empirical content — would misrepresent what has been done.

References

1. Stockman J, Fascetti AJ, Kass PH, Larsen JA. Evaluation of recipes of home-prepared maintenance diets for dogs. J Am Vet Med Assoc. 2013;242(11):1500–1505. doi:10.2460/javma.242.11.1500
2. National Research Council (US), Committee on Animal Nutrition. Nutrient Requirements of Dogs and Cats. Washington (DC): The National Academies Press; 2006. doi:10.17226/10668
3. FEDIAF. Nutritional Guidelines for Complete and Complementary Pet Food for Cats and Dogs. Brussels: FEDIAF; 2024. https://europeanpetfood.org/self-regulation/nutritional-guidelines/ (This is the edition on which the nutrient tables used here are based; a 2025 edition was published on 8 September 2025 and is not the basis of these runs.)
4. Association of American Feed Control Officials. Dog and cat food nutrient profiles. In: AAFCO Official Publication. Champaign (IL): AAFCO; 2016. (The 2016 profiles are the edition used by the model.)
5. Burger IH, Johnson JV. Dogs large and small: the allometry of energy requirements within a single species. J Nutr. 1991;121(11 Suppl):S18–S21. doi:10.1093/jn/121.suppl_11.S18
6. Kleiber M. Body size and metabolism. Hilgardia. 1932;6(11):315–353. doi:10.3733/hilg.v06n11p315
7. Turnlund JR, Keyes WR, Anderson HL, Acord LL. Copper absorption and retention in young men at three levels of dietary copper by use of the stable isotope ⁶⁵Cu. Am J Clin Nutr. 1989;49(5):870–878. doi:10.1093/ajcn/49.5.870
8. Billinghurst I. Give Your Dog a Bone. Bathurst (NSW): Warrigal Publishing; 1993. ISBN 0-646-16028-1
9. Fascetti AJ, Delaney SJ, editors. Applied Veterinary Clinical Nutrition. 1st ed. Ames (IA): Wiley-Blackwell; 2012. doi:10.1002/9781118785669. (A second edition was published in 2024; the first edition is the one consulted here.)
10. Delaney SJ. Balance IT [online home-diet formulation tool]. Davis (CA): DVM Consulting; 2003–. https://balanceit.com (accessed 7 September 2026).
11. Wilson EB. Probable inference, the law of succession, and statistical inference. J Am Stat Assoc. 1927;22(158):209–212. doi:10.1080/01621459.1927.10502953

How to cite

Prasanna M. (2026). Validating Residual Correction in Fresh-Food Companion-Animal Diets: A Paired-Arm Reliability Ablation, a Stochastic Reliability Framework, and a Tissue-Kinetic Safety Model (Version 3.0, Manuscript 8 of 8, Release B-2.1.0). Growlrr Foods Pvt Ltd. Zenodo. https://doi.org/10.5281/zenodo.22647623

The series

This manuscript is one of eight deposited together under release B-2.1.0, with a series cover note carrying the concept DOI and the full DOI list. All eight apply the BARC Protocol, published openly and separately at growlrr.com/barc-protocol.