{"id":"63efab77-cbb3-4da3-a706-05d2278e6bdc","arxiv_id":"1908.05339","paper_version":2,"verdict":"REJECT","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"high","formal_verification":"none","parameter_count":2,"one_line_summary":"A mixed hierarchical seasonality model estimates day-of-week and day-of-month parameters and combines them with a learned mixture weight for pallet transport forecasting.","lead":"Forecasting pallet volumes, the paper proposes a hierarchical Bayesian model that learns weekly and monthly seasonal effects separately and blends them with a learned weight. It reports accuracy gains on one logistics dataset, but its own tables undercut the headline claims.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Paper's own Table 1 shows MHS mixed pooling (0.95 MAPE) loses to Fourier (0.90) on the sole multiple-seasonality dataset, while Table 2 inverts the ELPD sign convention; the central empirical superiority claim is directly contradicted.","rationale":"The reader's strongest_claim is the empirical superiority of MHS, and their rationale explicitly notes that mixed pooling loses to Fourier in MAPE on the multiple-seasonality dataset and that the ELPD sign convention is inverted. My stress-test reaches the same conclusion: the paper's own tables contradict the abstract and Section 5.1. The reader's weakest_assumption focuses on the linear-mixture functional form; that is a legitimate modeling concern but not the most load-bearing issue here, because the reported empirical results are directly self-contradictory. I therefore agree with the rejection verdict but partially disagree with the emphasis: the decisive flaw is not an untested parametric assumption but the failure of the claimed empirical result under the paper's own evaluation. No further analysis is needed to see that the headline claim cannot be accepted as stated. The concrete test I propose would confirm the contradiction and add quantitative rigor (e.g., significance testing) if the authors or reviewers re-examine the results.","tokens_in":11674,"tokens_out":2038,"duration_ms":20764,"concrete_test":"Recompute Table 1 and Table 2 from the raw shipment forecasts across the 12 sliding windows: calculate MAPE for Fourier and mixed pooling, and compute ELPD using the loo package with the standard convention (higher is better). If mixed pooling still shows MAPE 0.95 vs. Fourier 0.90, or if the ELPD comparison reverses sign under the correct convention, the central claim of considerable improvement is unsupported. As a robustness check, run a paired Diebold-Mariano test on the windowwise absolute errors to assess whether any observed difference is statistically significant.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim is empirical: MHS achieves considerable improvements over Fourier, complete pooling, and SARIMA. The supporting evidence is in Tables 1 and 2. For shipment, the only dataset with multiple seasonalities where mixed pooling is appropriate, Table 1 reports MAPE: Complete pooling 1.43, Fourier 0.90, SARIMA 1.30, Mixed pooling 0.95. Thus mixed pooling is worse than Fourier by 0.05 MAPE, contradicting the claim of better performance. Table 2 reports ELPD: Fourier -1053, Mixed pooling -1279, but the text states 'Lower ELPD implies a better model.' ELPD is a predictive density measure where higher values are better; with the conventional orientation, mixed pooling is worse by 226 units. Either the sign convention is inverted or the models are mislabeled, and the stated interpretation is not valid. Moreover, the mixed pooling advantage is tested on a single proprietary dataset with no code or data release and no uncertainty quantification for the forecast error differences, so the claimed superiority is not established. The modeling idea of partial pooling for seasonal subcategories may have merit, but the central empirical claim fails on the paper's own reported numbers.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a Bayesian 'mixed hierarchical seasonality' (MHS) model for time series with multiple seasonal patterns. The model gives each day-of-week and day-of-month subcategory its own trend slope and intercept, partially pools those parameters through hyperpriors, and combines the weekly and monthly seasonal components via a simplex weight vector theta. The authors compare partial pooling (for single-seasonality series) and mixed pooling (for the series with both weekly and monthly seasonality) against complete pooling, Fourier decomposition, and SARIMA on three daily pallet-transport series from a logistics network. They report MAPE and ELPD as out-of-sample measures and claim that the proposed models achieve 'considerable improvements' over the baselines, especially for the shipment series, which exhibits both weekly and monthly seasonality.","tokens_in":11951,"tokens_out":5280,"duration_ms":49446,"significance":"If the reported improvements were reliable, the MHS model would offer a useful and interpretable alternative for business time series with strong, multiple seasonal cycles. The paper includes Stan code in Appendix A and applies the model to a real logistics dataset, which are positive features. However, the significance is seriously undermined by the fact that the paper's own tables contradict its central empirical claims: the mixed pooling model loses to Fourier decomposition on the one dataset where it is used, and the ELPD table appears to be interpreted with an inverted sign convention. As presented, the paper does not establish the claimed advantage of the proposed approach.","major_comments":[{"comment":"For the shipment dataset, the only dataset for which the mixed pooling model is used, the reported MAPE for mixed pooling is 0.95, while the Fourier decomposition model achieves 0.90. The text states that 'For data with multiple seasonality, shipment, our proposed mixed pooling model showed better performance,' which is directly contradicted by the table. Since the abstract's central claim of 'considerable improvements' over Fourier decomposition rests on this comparison, the main empirical result is not established.","section":"Section 5.1, Table 1"},{"comment":"The text states that 'Lower ELPD implies a better model,' but the loo package defines ELPD such that higher values indicate better out-of-sample predictive density. Under the correct sign convention, the table contradicts the paper's conclusions: for shipment, Fourier (-1053) beats mixed pooling (-1279); for restocking, Fourier (-962) beats partial pooling (-1409); and for delivery, complete pooling (268) beats both Fourier (-728) and partial pooling (-729). Thus the statement that 'Partial pooling and mixed pooling outperforms the other models for every dataset' is only true under the inverted sign convention and is false when ELPD is interpreted correctly.","section":"Section 5.1, Table 2"},{"comment":"The evaluation is based on a single proprietary dataset, and no data or code are released, so the results are not reproducible. Furthermore, the MAPE differences are reported without any uncertainty quantification across the 12 sliding-window test sets; for example, the delivery comparison (31.17 vs 31.33) and the shipment comparison (0.95 vs 0.90) could easily be within sampling noise. The choice of pooling structure (partial vs mixed) is made after inspecting the same data, which introduces a selection effect that is not accounted for in the reported comparisons.","section":"Section 4, Section 5.1"},{"comment":"The mixed pooling model assumes that the combined weekly and monthly seasonal effect is a linear mixture of separately estimated slope and intercept parameters with a single global weight theta. This linearity assumption is neither justified nor tested, and the shipment dataset—the only place where this assumption is exercised—shows the model underperforming the Fourier baseline. This outcome is consistent with the possibility that the interaction between the two seasonal patterns is not a simple weighted average, which is a fundamental misspecification concern for the proposed model.","section":"Section 3.3, Eq. (4c)"}],"minor_comments":[{"comment":"Equation (3b) appears twice; the second displayed equation should be numbered (3c) or later equations should be renumbered.","section":"Section 3.2"},{"comment":"The sentence 'The former three correspond to' is incomplete; the SARIMA parameter sets are listed but the clause is not finished.","section":"Section 4"},{"comment":"The word 'Mulitseasonal' should be 'multiseasonal'.","section":"Section 2.1"},{"comment":"The caption states that 'The mixed pooling model shows a significantly higher performance than the other models,' but Figure 4 displays restocking data, which was analyzed with partial pooling, not mixed pooling; this is misleading.","section":"Figure 4 caption"},{"comment":"Several cells contain stray periods (e.g., '31.17 (week pool) .'), which appear to be formatting artifacts and should be removed.","section":"Table 1"},{"comment":"The citation 'Hyndman, 2018' in the introduction appears to refer to Hyndman and Athanasopoulos (2018), but the reference list contains only the latter; the in-text citation should be corrected.","section":"Introduction"}],"recommendation":"reject","confidential_remarks":"The manuscript's own tables contradict its central claims, and the ELPD sign error suggests a fundamental misreading of the loo package output. The empirical basis for the paper is therefore unsound as submitted; even if the ELPD issue were a simple sign reversal, the MAPE results still show mixed pooling losing to Fourier on the shipment dataset. Substantial new experiments on additional datasets and a corrected analysis would be needed before the claims could be considered, so rejection is appropriate. The authors might consider a resubmission after major revision, but the current version does not meet the bar for publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The modeling idea is coherent, but the paper's central empirical claim fails on its own numbers. On the only multiple-seasonality dataset, the mixed model's MAPE (0.95) is worse than Fourier's (0.90), and the ELPD comparison is reported with an inverted sign convention.\n\nWhat's new: a hierarchical partial-pooling setup for day-of-week and day-of-month effects, blended by a simplex weight, with Stan code in the appendix. For single-seasonality datasets, partial pooling does help—restocking MAPE drops from 13.09 (Fourier) and 7.00 (SARIMA) to 4.09. The theta posterior is a nice interpretable output for relative seasonal strength.\n\nWhere it falls apart: shipment is the dataset where mixed pooling is supposed to shine, but Table 1 shows it loses to Fourier. Table 2 says 'Lower ELPD implies a better model' but ELPD is higher-better; with the loo package, the mixed model's -1279 is worse than Fourier's -1053. That is not a typo—it inverts the conclusion. Add to that a single proprietary dataset, no code or data release, no uncertainty on the error differences, and a model structure (linear mixture) that is never checked against alternatives. The pooling structure is also chosen per dataset after inspecting the same data, so there is a mild selection effect.\n\nWho it's for: applied forecasters wanting a simple Bayesian seasonal decomposition. But as it stands, the evidence doesn't support the headline. The partial pooling variant might have merit if tested properly on public data with correct error metrics. I wouldn't cite it, and I'd desk reject: the central claim is contradicted by the paper's own tables, so referee time would go to fixing fundamentals rather than polishing.","headline":"The modeling idea is coherent, but the paper's central empirical claim fails on its own numbers: the mixed model loses to Fourier on the only multiple-seasonality dataset, and the ELPD comparison is sign-inverted.","tokens_in":12424,"tokens_out":2797,"would_cite":false,"duration_ms":25696,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A mixed hierarchical seasonality model learns each seasonality's weight from data and forecasts pallet transport more accurately than complete pooling, Fourier decomposition, or SARIMA.","keywords":["multiple seasonality","hierarchical model","partial pooling","mixed pooling","time series forecasting","Bayesian inference","pallet transport","forecast accuracy"],"falsifier":"A decisive check is already in the paper's Table 1: for the shipment flow, the mixed pooling MAPE (0.95) is worse than Fourier decomposition (0.90), so a replication on additional multi-seasonal datasets, with uncertainty intervals around MAPE, would settle whether the claimed superiority over Fourier is real; a complementary simulation with a multiplicative weekday-month interaction would test whether the linear-mixture assumption is identifiable.","tokens_in":11484,"feed_emoji":"📦","tokens_out":15212,"duration_ms":138541,"temperature":0.7,"pith_summary":"The paper argues that a time series with several seasonal rhythms should not be forced through one global trend plus additive seasonal terms. It introduces the mixed hierarchical seasonality (MHS) model, in which each day of the week and each day of the month gets its own slope and intercept drawn from shared Bayesian priors, and the forecast is a weighted average of these per-period parameters with weights learned from the data. On three pallet-transport flows, it reports out-of-sample forecast-error reductions against complete pooling and SARIMA, and better predictive density than other Bayesian models. The practical stake is that forecasts of demand and transport volume become both more accurate and more interpretable, because the learned mixture weight says which calendar rhythm actually dominates.","feed_headline":"Mixing weekday and month-day trends improves pallet forecasts","feed_subtitle":"The model learns whether weekly or monthly rhythm dominates, so demand forecasts are both sharper and easier to explain.","key_machinery":"The engine is the mixed pooling forecast, in which the slope and intercept at time $t$ are $\\sum_{d=1}^D \\theta_d\\, k_{d,\\,j_d(t)}$ and $\\sum_{d=1}^D \\theta_d\\, m_{d,\\,j_d(t)}$, where $d$ indexes the seasonal patterns (week, month), $j_d(t)$ is the subcategory (weekday or month-day) of time $t$ in pattern $d$, and the parameters $k$ and $m$ are drawn from shared normal hyperpriors. The hierarchical priors give partial pooling, so a distinctive period such as Sunday or month-end keeps its own shape while borrowing strength from other periods. The $\\theta$ vector is a simplex, with components summing to one, and is learned from the data, so the model automatically sets the relative importance of each seasonality. This weighted-average construction, rather than an additive sum of seasonal effects, is what gives the model its name and its claimed ability to handle interactions.","core_discovery":"The central claim is that seasonal interactions can be learned rather than pre-specified. In the mixed pooling model, subcategory parameters $k_{ij}$ and $m_{ij}$ stand for the slope and intercept of day-of-week or day-of-month $j$ within seasonality $i$, sharing normal hyperpriors; the forecast mean is a weighted average of those per-period vectors with a simplex weight $\\theta$. This is the mechanism that lets the model represent the combined effect of week and month as a learned mixture instead of an assumed sum. On the delivery and restocking flows, which have a single dominant season, the paper reports that partial pooling cuts MAPE sharply (delivery from 292.13 to 31.17; restocking from 12.66 to 4.09). On the shipment flow, where weekly and monthly rhythms interact, it reports that mixed pooling performed best and that the posterior weight, roughly $\\theta_{\\text{week}} = 0.6$ and $\\theta_{\\text{month}} = 0.4$, identifies weekly seasonality as the larger contributor.","pith_inferences":["The per-period posterior distributions double as a diagnostic for anomalous calendar days; for instance, the paper's own parameter plots single out Sunday in delivery and month boundaries in restocking, which could be used to flag calendar effects the model cannot explain.","If the linear mixture is correct, then the mixture weight $\\theta$ should be stable across rolling windows for a stationary seasonal regime; a drifting $\\theta$ would be a simple, testable early warning for changing seasonal behavior.","A natural extension is to treat $\\theta$ as a group-level parameter shared across related series, such as all flows in a logistics network, shrinking individual weights toward a common seasonal profile and borrowing strength when one flow has little data."],"forward_implications":["For single-seasonality series, partial pooling alone captures periodic scale jumps; the delivery MAPE drops from 292.13 (complete pooling) to 31.17 (week pooling).","For multi-seasonality series, forecasters no longer need to guess which season dominates: the model learns the relative strength from data, unlike SARIMA or Fourier setups that require the seasonal form to be fixed in advance.","The learned $\\theta$ gives an operational handle: managers can rank weekly versus monthly effects and time replenishment or staffing decisions accordingly.","Because the model only assumes an interchangeable hierarchy, the same construction extends to hourly, daily, quarterly, or yearly seasonalities by adding dimensions to the parameter arrays and the simplex."],"supporting_citations":[{"why":"Supplies the theoretical basis for hierarchical partial pooling: sharing information across subcategories almost always improves estimates, the claim the MHS model extends.","marker":"Gelman (2006a)"},{"why":"Defines partial pooling and Bayesian hierarchical models, the statistical machinery used in equations (3a)-(4c).","marker":"Gelman et al. (2013)"},{"why":"Supplies the Fourier-decomposition baseline (Prophet-style seasonality) against which MHS is compared and motivates the need to model multiple seasonalities.","marker":"Taylor and Letham (2018)"},{"why":"Provides the time-series cross-validation protocol (sliding windows), the MAPE accuracy measure, and the critique of existing multiple-seasonality methods.","marker":"Hyndman and Athanasopoulos (2018)"},{"why":"Supplies the Pareto smoothed importance sampling and ELPD used to compare predictive density of the Bayesian models.","marker":"Vehtari et al. (2017)"},{"why":"Provides the probabilistic programming language used to implement and sample the pooling models.","marker":"Carpenter et al. (2017)"},{"why":"Prior application of season-based clustering to pallet movement data, which motivates grouping calendar periods and whose hard-cluster structure the mixed pooling model aims to soften.","marker":"Moon and Song (2019)"}],"fun_headline_variants":["Learned blend of weekly and monthly rhythms cuts pallet forecast error","Pallet forecasts sharpen when weekly and monthly trends are mixed","Mixed pooling adapts weekly and monthly weights for pallet demand","Model learns to mix weekday and month-day seasons for pallet forecasts","Weekly and monthly rhythms combined reduce pallet prediction error"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The model treats the interaction between weekly and monthly seasonality as a single weighted average of separately estimated trend lines, with one global weight per season applied to both slope and intercept, so if the true interaction is not that simple linear mixture, the forecasts are structurally wrong; the paper tests this on only one shipment dataset.","fun_headline_variants_meta":{"raw":{"variants":["Learned blend of weekly and monthly rhythms cuts pallet forecast error","Pallet forecasts sharpen when weekly and monthly trends are mixed","Mixed pooling adapts weekly and monthly weights for pallet demand","Model learns to mix weekday and month-day seasons for pallet forecasts","Weekly and monthly rhythms combined reduce pallet prediction error"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000828,"raw_usage":{"total_tokens":3606,"prompt_tokens":919,"completion_tokens":2687,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":535,"completion_tokens_details":{"reasoning_tokens":2602}},"tokens_in":535,"tokens_out":2687,"duration_ms":21293,"temperature":1.0,"reasoning_tokens":2602,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:17:10.248059+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A decisive check is already in the paper's Table 1: for the shipment flow, the mixed pooling MAPE (0.95) is worse than Fourier decomposition (0.90), so a replication on additional multi-seasonal datasets, with uncertainty intervals around MAPE, would settle whether the claimed superiority over Fourier is real; a complementary simulation with a multiplicative weekday-month interaction would test whether the linear-mixture assumption is identifiable.","supporting_citations":[],"review_version":1}