{"id":"2c07b69a-be6d-4335-9376-0a4883d778cb","arxiv_id":"2607.21741","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Aligned PCA adds a low-rank, tunable penalty to each group's covariance matrix to rotate group-specific principal components toward the pooled-data principal directions, trading a little explained variance for cross-group comparability.","lead":"This paper introduces 'aligned PCA,' which nudges each group's principal components toward the directions found in the pooled data by adding a tunable penalty to each group's covariance matrix. It is a simple method for multi-group datasets that aims to keep within-group variation while making components comparable across groups.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central claim is untested when pooled directions are dominated by between-group mean separation: simulations set all group means to zero, and the census results show large PVE losses for the least-aligned group.","rationale":"The paper's construction is transparent, and the mathematical claims (Propositions 1–2) appear sound; the method is a sensible regularized compromise and the code is promised. However, the central empirical claim—that aligned PCA achieves a favorable variance/alignment trade-off—is established only in a zero-mean simulation design. The reader's concern about pooled directions being dominated by between-group mean differences is well-founded: §2.1 explicitly says the pooled covariance combines within-group and between-group variation, but the simulation removes the latter by generating all groups from N(0, Σ_g). Real data have nonzero group means, and the census application shows that the least aligned region (Atlantic) pays a large PVE price when rotated toward the national subspace. This is not a mathematical inconsistency; it is an untested regime in which the compromise may fail. A controlled simulation with mean separation orthogonal to the common within-group signal would settle whether the method can still deliver the claimed compromise. The reader's conditional verdict already captures the need for such evidence, so I would not change it.","tokens_in":18404,"tokens_out":6054,"duration_ms":57085,"concrete_test":"Simulate G=8 groups, p=10, r_true=3 with the same eigenvalue structure as §3, but set group means μ_g = c·u_perp, where u_perp is orthogonal to the shared directions u_0,1,...,u_0,3 and c is chosen so that the between-group variance along u_perp equals λ1=4 (i.e., roughly the leading within-group eigenvalue). Generate X_ig ~ N(μ_g, Σ_g), estimate the pooled covariance, and run aligned PCA at τ=0, 0.5, 1, 2, 3 with uniform weights. Compute PVE, A, and S as in §2.4, and also the squared cosine between the leading pooled eigenvector bv_1^global and the span of {u_0,1,...,u_0,3}. If, at moderate τ, the aligned PVE loss exceeds, say, 10% relative to group-wise PVE while the same τ in the zero-mean design loses <5%, the favorable-compromise claim is not robust to mean-separation-dominated global directions. Equivalently, repeat Table 2 with this one change; the PVE column for aligned PCA is th","verdict_should_be":"UNCHANGED","load_bearing_attack":"The method's target V_r is the pooled eigenbasis of Σ (Eq. 1, §2.1). The paper itself notes that Σ \"combines the within-group covariance structure and the between-group differences when the group means are not identical\" (§2.1). Yet the simulation study (§3) generates X_ig ~ N(0, Σ_g), so group means are identical and the pooled directions contain no between-group component. The favorable-compromise claim — that moderate τ preserves within-group variance while substantially improving alignment — is therefore only demonstrated in a regime where the global directions are, by construction, common within-group directions. In the census application, regional means differ, so pooled V_r can be dominated by between-region contrasts. The region-level results show the cost of alignment is not uniformly modest: for Atlantic Canada at r=3, τ=3, PVE drops from 79.07% to 55.16%, and for the first component from 34.52% to 28.57% while its angle to the national PC falls from 74.7° to 19.2° (Tables 5 and 7). This is the signature of a global direction carrying little within-group variance: reaching alignment forces a large PVE sacrifice. The paper never checks whether V_r is near the within-group subspace, nor does it investigate mean-shift heterogeneity. Without such a check, the headline \"favorable compromise\" is conditional on the pooled directions not being driven by group-mean separation.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a globally aligned PCA for multi-group data. For each group, it adds a rank-regularizing term to the group covariance matrix, constructed from the leading eigenvectors of the pooled covariance matrix; a scalar parameter τ controls how strongly group-specific principal components are pulled toward those global directions. Two propositions characterize the limiting behavior of the aligned components and of their within-group explained variance as τ→∞. The method is illustrated in a simulation study and applied to 2021 Canadian Census data, where it is claimed to achieve a favorable compromise between preserving within-group variance and improving cross-group alignment and stability.","tokens_in":18820,"tokens_out":9328,"duration_ms":93784,"significance":"The construction is clean and computationally tractable, and the explicit limiting results (Propositions 1–2) are useful for users who want to understand the effect of the tuning parameter. The paper also provides reproducible R code. If the claimed favorable compromise holds in realistic multi-group settings, the method could be a practical contribution for constructing comparable area-based indices and for other multi-group PCA applications. However, the central empirical claim is currently established only under a restricted zero-mean simulation design, and the main evaluation indices are partly constructed from the same global directions that the penalty targets. As it stands, the significance is therefore conditional on the missing analyses below.","major_comments":[{"comment":"The simulation study tests the method only when all group means are equal to zero. As the manuscript notes in §2.1, the pooled covariance Σ includes between-group mean differences when group means are not identical. Yet the simulation generates X_ig ∼ N(0,Σ_g), so the pooled eigen-directions V_r contain no between-group component. The favorable-compromise claim is therefore demonstrated only in the regime where V_r is a common within-group direction. The Census application is the only place with nonzero mean differences, and there the trade-off is not uniformly favorable: Table 5 (r=3, τ=3) shows Atlantic Canada's PVE falling from 79.07% to 55.16% while A_g rises from 0.690 to 0.988, and Table 7 shows its PC1 PVE falling from 34.52% to 28.57% while the angle to the national PC drops from 74.7° to 19.2°. This is the signature of a pooled direction carrying little within-group variance. Pl","section":"§3 and §2.1"},{"comment":"The main empirical evidence is based on indices defined relative to the same subspace used in the penalty. The alignment index A is the average squared cosine between the group-aligned subspaces and V_r^global, the very matrix inserted into Σ_g^(ρ) in Eq. (1). Proposition 1 shows that with strictly ordered weights, V_g^(ρ)→V_r^global as τ→∞, so A→1 and S→0 are consequences of the regularization rather than independent discoveries. The paper should treat A and S as convergence diagnostics and support the 'favorable compromise' claim with a direct analysis of the PVE loss per unit gain in A/S, or with a comparison against a simple shrinkage-to-global baseline with the same number of parameters. Without that, the reported improvements in alignment and stability in Tables 1–4 are partially circular.","section":"§2.4, Eq. (5); Tables 1–4"},{"comment":"All Monte Carlo results are reported as point averages over 200 replications, without Monte Carlo standard errors, confidence intervals, or any measure of dispersion. This matters because some of the headline comparisons are between quantities that differ by only a few percent (for example, Table 3 shows A for the aligned method increasing from 0.86 at r=1 to 0.93 at r=3). The reader cannot judge whether such differences are systematic or sampling noise. Please provide standard errors, boxplots, or replication-level summaries for W, PVE, A, and S.","section":"§3, Tables 1–3"},{"comment":"The guided-selection rule defines τ*(ϵ) as the largest τ such that PVE_g(V_g^(ρ)) ≥ (1−ϵ) PVE_g(V_g) for every group. This presumes that PVE_g(V_g^(ρ)) is monotone nonincreasing in τ. Since V_g^(ρ) is the leading eigenspace of a rank-perturbed covariance matrix, it is not obvious that PVE under the original covariance Σ_g is monotone, especially when eigenvalues cross or are nearly degenerate. If monotonicity fails, a grid search over [0,5] may return a local rather than the global maximum, and Table 6 may not give the stated 'largest' τ. Please provide a proof or a numerical check, for example by plotting PVE_g as a function of τ for each region in the Census application.","section":"§2.3, Eq. (2)"}],"minor_comments":[{"comment":"Proposition 1 requires strictly ordered weights w1 > ... > wr, while §2.3 recommends uniform weights as a simple choice. The paper notes that individual directions do not converge under uniform weights, but it would be helpful to state the practical implication for interpreting the aligned components in that case.","section":"§2.5"},{"comment":"The 'stability index' S is defined so that S=0 is perfect stability and S=1 is the worst case. Consider calling it an instability index, or add a sentence making the orientation explicit, to avoid confusion.","section":"§2.4, Eq. (6)"},{"comment":"The proofs of Propositions 1 and 2 are deferred to a Supplementary Material that was not available to the referee. For a journal submission, the proofs should be included in the manuscript or in the review package.","section":"§2.5"},{"comment":"The notation PVE_g/A_g and PVE_a/A_a is easy to confuse; use explicit subscripts such as PVE_groupwise, PVE_aligned, etc., or a clear table caption.","section":"§4.1, Tables 5 and 7"},{"comment":"Figure 1 is described as an animation. In the static PDF version, provide static snapshots for several τ values or an external link, so the trade-off is visible in print.","section":"§2.2"},{"comment":"Typo: 'We aims to improve' should be 'We aim to improve'.","section":"§2.3"}],"recommendation":"major_revision","confidential_remarks":"The core idea is simple and likely useful, but the central claim is currently conditional on zero group means and on evaluation metrics that are partly built into the construction. I would be willing to accept after the authors add a mean-shift simulation, quantify the PVE/alignment trade-off independently, and report Monte Carlo uncertainty. The method itself is defensible; the missing work is within the scope of a revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Globally aligned PCA is a modest but honest addition to the multi-group PCA toolbox. The construction—Σ_g + V_r D_ρ V_r^T—is simple and computationally trivial, and the paper is upfront that it interpolates between group-wise and pooled PCA. The limiting propositions are plausible; proofs live in the supplement, so I'd want those checked rather than taken on faith. The R code is public and the census application is real, which helps reproducibility.\n\nWhat is genuinely new is packaging that low-rank perturbation as a tunable alignment framework, with separate control over the number of global directions r and the alignment strength τ, plus practical guided selection of τ via a PVE-loss tolerance. That is a useful distinction from CPC, JIVE, and RGCCA, and the paper cites that literature fairly.\n\nThe soft spots are not fatal, but they are real. First, the simulation never touches the regime the paper itself identifies in §2.1: pooled covariance mixes within-group and between-group variation when group means differ. The simulation draws all groups from N(0, Σ_g), so the pooled directions are, by construction, common within-group directions. The real census regions have different means, and Tables 5 and 7 show the consequence: Atlantic Canada loses 24 points of PVE at r=3, τ=3 while its angle to national PC1 falls from 74.7° to 19.2°. That's the signature of a global direction that carries little within-group variance, and the paper never checks whether V_r is close to the within-group subspace. The favorable-compromise claim is therefore conditional on a regime the paper's own main application is not in.\n\nSecond, the evaluation is partly circular. The alignment index A is squared cosine to the same V_r that the penalty inserts, so increasing τ is mechanically expected to raise A; Proposition 1 makes that explicit. That does not make the method worthless—the PVE loss is the real constraint—but it does mean the headline gain in A is built into the objective. The paper would be much stronger with an external baseline (CPC, JIVE, RGCCA, or at least a subspace-distance to an independent target) and with standard errors or MC intervals around the reported averages.\n\nWho is this for? Applied statisticians working with region/center/group-structured data who want a single tunable compromise between local fit and a pooled reference. It is not a breakthrough, but it is a workable tool. It deserves a serious referee; the revision should add a mean-shift simulation, a diagnostic for how much of V_r lies in the within-group space, an external method comparison, and error bars. Send it to review, but expect the mean-shift issue to drive the revision.","headline":"A simple, tunable low-rank penalty for multi-group PCA that is worth engaging, but the paper's headline compromise is only demonstrated in a mean-zero simulation regime and partly measured by a circular alignment index.","tokens_in":19234,"tokens_out":2965,"would_cite":true,"duration_ms":33967,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["62H25"],"pacs":[],"model":"deepseek-v4-flash","headline":"A new PCA variant for grouped data adds a low-rank penalty to each group's covariance, rotating its principal components toward the pooled data's global directions while preserving most within-group variance.","keywords":["globally aligned PCA","multi-group data","principal component analysis","covariance regularization","global alignment","subspace stability","dimension reduction","common principal components"],"falsifier":"Simulate several groups with identical within-group covariance but far-apart means; if the first aligned component under large tau converges to the direction separating the group centroids rather than to the common within-group major axis, the alignment target is encoding between-group differences, not shared structure.","tokens_in":18346,"feed_emoji":"🧭","tokens_out":3742,"duration_ms":35717,"temperature":0.7,"pith_summary":"The paper proposes a principal component analysis for multi-group data that occupies the middle ground between pooling all observations and analyzing each group separately. The central idea is to add a low-rank penalty to each group's covariance matrix along the top pooled (global) principal directions, so that each group's leading components rotate toward a common frame while retaining most of the group's own variance. The strength of this rotation is controlled by a single parameter tau, with tau=0 giving ordinary group-wise PCA and large tau reproducing the pooled solution. Simulations and an application to Canadian Census socioeconomic data show that moderate tau substantially raises alignment and cross-group stability at a small cost in explained variance.","feed_headline":"New PCA variant aligns group components with a global frame","feed_subtitle":"Adding weighted global directions to each group's covariance preserves local variance while making components stable across groups.","key_machinery":"The globally aligned covariance matrix: each group covariance sigma_g is perturbed by a weighted sum of outer products of the r leading pooled principal directions, with weights rho_m = tau w_m. Geometrically, the perturbation inflates variance along the global directions in proportion to the squared cosine of the angle to those directions, making them more likely to appear among the group's leading eigenvectors. This low-rank penalty is what carries the alignment mechanism.","core_discovery":"The discovery is that the globally aligned covariance matrix, sigma_g^(rho) = sigma_g + sum_{m=1}^r rho_m v_m^global (v_m^global)^T, defines a simple, closed-form family of intermediate methods. For any group, the leading eigenvectors of this modified covariance are the group's aligned components. As tau grows, these components converge, up to sign, to the global directions, while their within-group explained variance converges to the variance the global direction captures in that group; at tau=0 they coincide with the group-wise components. This yields an explicit interpolation with a tunable trade-off between local fit and global comparability.","pith_inferences":["The paper targets alignment with pooled principal directions, but the same mechanism could be pointed at any reference subspace; a natural variant would align groups to a within-group common subspace (e.g., a robust or covariance-pooled estimate) to avoid confounding with between-group mean differences.","The explicit trade-off suggests a testable rule for practice: choose the smallest tau that meets a stability criterion, since the paper's own tables show the variance cost rises steeply only after large gains in alignment are already captured.","The framework's geometric construction extends directly to kernel or functional PCA, as the authors note; one could align group eigenfunctions by adding a rank-one operator along pooled eigenfunctions.","If the goal is cross-group comparability of scores, aligning loadings is only half the story; computing scores in the aligned basis and checking their distributions across groups would be a direct downstream validation."],"forward_implications":["At moderate alignment strengths, the method achieves a large fraction of the possible gain in global alignment and stability while losing only a few percentage points of within-group explained variance.","The method contains group-wise PCA (tau=0) and, in the limit, the pooled solution as endpoints, so it offers a continuous bridge rather than a new disjoint estimator.","Users can select tau to respect a maximum per-group PVE loss; even a 1% tolerance already improves alignment substantially in the census example.","Regions that are already close to the global subspace are barely changed, while atypical groups are rotated the most—alignment acts as a gentle regularizer, not a forced pooling.","Because the aligned components are just eigenvectors of a modified covariance, the method inherits the computational simplicity of standard eigendecomposition."],"fun_headline_variants":["Aligned PCA blends local and global structure for multi-group data","Tunable PCA method balances group-specific and global components","New PCA with aligned components improves cross-group comparability","Globally aligned PCA offers explicit trade-off between local and global","Multi-group PCA with weighted global directions for stable components"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The global directions come from the pooled covariance around the overall mean, so the method assumes those directions reflect a shared within-group structure rather than being dominated by differences between group averages—if that assumption fails, rotating groups toward them could replace local variation with between-group contrast.","fun_headline_variants_meta":{"raw":{"variants":["Aligned PCA blends local and global structure for multi-group data","Tunable PCA method balances group-specific and global components","New PCA with aligned components improves cross-group comparability","Globally aligned PCA offers explicit trade-off between local and global","Multi-group PCA with weighted global directions for stable components"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000155,"raw_usage":{"total_tokens":1026,"prompt_tokens":696,"completion_tokens":330,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":440,"completion_tokens_details":{"reasoning_tokens":262}},"tokens_in":440,"tokens_out":330,"duration_ms":3683,"temperature":1.0,"reasoning_tokens":262,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T06:49:04.099781+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate several groups with identical within-group covariance but far-apart means; if the first aligned component under large tau converges to the direction separating the group centroids rather than to the common within-group major axis, the alignment target is encoding between-group differences, not shared structure.","supporting_citations":[],"review_version":1}