{"id":"f3395e7e-e1d2-40f8-a60a-0b80c10297f9","arxiv_id":"2607.20696","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"CEDAR recovers lagged cross-variable causal edges in sparse AR(1)-like time series using AR(1)-residualized distance-correlation screening, two CI tests per candidate lag, and stable MCI pruning.","lead":"A new method finds delayed causal links between time-series variables by combining distance-correlation screening with targeted conditional-independence tests and a pruning step. It targets small-sample autoregressive problems where richer methods lose statistical power.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1's oracle guarantee is near-tautological: A6 and A7 are stated in terms of CEDAR's own Phase-2 outputs, so the theorem assumes the very success it claims to prove.","rationale":"The reader correctly identified A7 as a fragile assumption; my stress-test sharpens this to a more basic concern: A7 is not a graph-structural condition but a condition on CEDAR's own Phase-2 outputs, and A6 is essentially the algorithm's success criterion. This does not make the empirical method useless — the ablation study, stress tests, and real-world Elbe result are genuine evidence of practical utility — but it does mean the oracle guarantee is much weaker than the phrase 'exactly recovers' suggests. The paper is honest about A6/A7 being strong conditions, yet the theorem's framing still overstates the theoretical foundation. The stress-test discrepancy (5 injected A7 bypasses but only 0.6 FNs) warrants a concrete check rather than an immediate rejection. Since the reader's verdict was already CONDITIONAL, this concern reinforces that conditionality without moving the verdict.","tokens_in":20043,"tokens_out":14108,"duration_ms":114499,"concrete_test":"Run CEDAR's Phase 2 with oracle CI (d-separation queries on the true time-unrolled DAG) on a sample of Figure 1's SCPs and on the Appendix counterexample. For each true eligible edge, compute whether A7 holds using the Phase-2-selected \\hat w_{k,e}, and stratify edge recall by A7 status. If A7-violating true edges are not systematically missed, Theorem 1's necessary condition is not actually operative; if A7 violations are rare in the generator, the empirical claim is confined to graphs that satisfy the theorem's near-tautological assumptions.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central theoretical claim reduces to a near-tautology. A7's blocking set Z2 in Eq. (A-CB) contains S_{e,c} = {X_k^{t-\\hat w_{k,e}-1}}, where \\hat w_{k,e} are lags that CEDAR itself selects in Phase 2. A7 is therefore not a property of the true graph alone: whether a true edge X_c^{t-w}->X_e^t is 'blockable' depends on whether every other source-target pair was already recovered correctly. A6 adds that no candidate before the true lag passes Conditions 1-2, which is exactly the Phase-2 correctness the theorem is meant to establish. The Appendix's counterexample shows A1-A6 and A8 can hold while A7 fails, but it does not resolve this circularity: one false positive for another pair can alter \\hat w_{k,e}, destroy A7 for a true edge, and produce a missed edge even in a graph with perfectly aligned true lags. As a result, Theorem 1 does not provide a non-trivial structural condition under which CEDAR works; it states a consistency condition on CEDAR's own intermediate outputs. The empirical stress test in Table 9 ('A7 bypass×5') reports only 0.6 FNs despite introducing five bypass paths, which is difficult to reconcile with the guarantee that a single A7-violating bypass on a true edge should cause a missed edge — suggesting either the injected bypasses do not actually violate A7 or the oracle theory is not predictive of the finite-sample experiments.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes CEDAR, a constraint-based lagged causal edge discovery method for sparse autoregressive time series. It combines AR(1)-residualized distance-correlation lag screening, SYPI-style two-condition CI tests per candidate pair, and an order-independent MCI pruning step, with optional deterministic C-nodes for trend nonstationarity. The paper reports broad experiments: synthetic SCPs up to d=75, ablations, assumption-violation stress tests, a real-world Elbe river benchmark, and an oracle exactness theorem in Appendix A under assumptions A1–A8. The central empirical claim is that CEDAR is competitive or best in low-sample regimes (T≤200) while remaining interpretable and cheap; the formal guarantee is explicitly conditional and the paper acknowledges its method-specific assumptions.","tokens_in":20440,"tokens_out":6696,"duration_ms":59581,"significance":"If the empirical results hold, CEDAR is a useful incremental contribution for scarce-data causal discovery in sparse AR(1)-like systems. The paper's strengths are concrete: reproducible code, independent synthetic ground truth with exact-lag evaluation, a thoughtful ablation that identifies MCI pruning and AR(1)-residualized dcor as load-bearing, honest t-test calibration (Appendix C), and a real benchmark with a clear nonstationarity story. However, the theoretical appendix does not provide a graph-theoretic characterization of when CEDAR succeeds: assumptions A6 and A7 are stated in terms of CEDAR's own intermediate outputs, making Theorem 1 close to a consistency statement rather than a substantive structural guarantee. The empirical core is sound enough to merit publication after the theoretical framing and one stress-test experiment are revised.","major_comments":[{"comment":"Theorem 1's guarantee is near-tautological as stated. A6 requires that every candidate before the true lag fails Conditions 1–2 under oracle CI, which is exactly the Phase-2 first-accept success condition the theorem claims to establish; A7's conditioning set Z2 in Eq. (A-CB) contains the Phase-1-selected lags \\hat w_{k,e}, so whether a true edge is 'blockable' depends on lag choices made by CEDAR itself. The Appendix counterexample shows A1–A6 and A8 can hold with A7 failing, but it does not repair the issue: one incorrect lag choice for another pair can alter Z2 and destroy A7 for a true edge. The theorem should be re-framed as a conditional consistency statement, or A6/A7 should be replaced by sufficient conditions on the true graph (e.g., lag-alignment/order restrictions) that imply them.","section":"Appendix A, Theorem 1 (A6/A7)"},{"comment":"The A7 bypass stress test is not presently informative. By Prop. 1(ii), a true edge whose deeper source is not d-separated by Z2 should be missed under oracle CI. Table 9 reports only 0.6 FNs after introducing five bypass paths, which is difficult to reconcile with the theorem's failure mode. The paper does not state how the five bypasses are constructed, how many true edges they affect, or whether they actually violate A-CB with the Phase-1-selected lags. Without this, the experiment cannot support the 'graceful degradation' claim in the A7 dimension. Please specify the construction, verify at oracle level that A7 is violated, and report per-seed TP/FP/FN, or replace with a graph from Appendix A's counterexample.","section":"Table 9 / Section 4.2, A7 bypass×5"},{"comment":"The meaning of \\hat w_{k,e} is ambiguous and load-bearing. The main text and Appendix A use it to build S_c and Z2, but it is not specified whether \\hat w_{k,e} is the Phase-1 screened lag (highest dcor), the first lag accepted after Phase 2, or the selected lag after BH screening. Since Phase 2 processes sources sequentially, the 'other candidate' lags may not yet be finalized when S_c is built. This ambiguity affects both the executability of Algorithm 1 and the well-posedness of Theorem 1. Please define a deterministic order and state explicitly which lag is used in S_c/Z2; the formal proof should be written against that definition.","section":"Algorithm 1 line 6 and Section 3.2"}],"minor_comments":[{"comment":"The calibration table reports type-I error only under d=5 AR(1) nulls with no cross-links. A sentence noting that the analytic t-approximation is high-dimensional in origin and that calibration under cross-links or nonlinear nulls is future work would be helpful.","section":"Appendix C"},{"comment":"The abstract says CEDAR requires O(d^2) CI tests after screening; the body correctly notes the worst case is O(d^2 L_sig). The abstract's 'in sparse regimes' qualifier should be repeated whenever the O(d^2) complexity is cited.","section":"Section 3.6 / Abstract"},{"comment":"The Elbe result uses PELT segmented regimes with different lag bounds and an OR aggregation. It would help future readers to state whether the per-regime graphs are assumed to share the same variable set and how the union handles conflicting lags between regimes.","section":"Section 4.4 / Appendix J"},{"comment":"The claim that all Z1 variables are non-descendants of the target is correct, but the paragraph could note explicitly that conditioning on earlier variables that are colliders can still open paths; the text does say this, but a pointer to A7 would improve clarity.","section":"Appendix A, Sink-node relaxation paragraph"}],"recommendation":"major_revision","confidential_remarks":"The empirical core is honest and reproducible, and the method is a reasonable contribution to low-sample time-series causal discovery. The main obstacle is the theoretical appendix: as written, Theorem 1's A6/A7 assumptions are method-specific success conditions, so the 'oracle exactness' claim overstates what has been proven. This is fixable by re-framing the theorem as a conditional statement and by adding a genuine structural condition or a clearly labeled non-goal. I would also ask the authors to make the A7 stress test well-defined; the current result does not engage with the theorem's actual failure mode."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth your time. CEDAR is a practical method for lagged causal discovery in sparse, data-scarce time series, and the empirical work is the real substance. The core combination—AR(1)-residualized U-centered distance correlation for screening, SYPI-style two-condition tests extended to all ordered pairs, stable MCI pruning, and optional C-nodes—is a sensible engineering contribution. The ablations are honest and informative: they isolate the contribution of each piece, show that MCI pruning and AR(1)-residualized dcor carry most of the weight, and acknowledge the non-monotone behavior of intermediate steps. The Elbe river benchmark is a genuine plus: with PELT segmentation and per-regime runs, CEDAR recovers 10 of 11 edges, and the authors correctly identify the low-flow regime as the cleanest signal. That is reproducible, falsifiable evidence.\n\nThe stress-test note is right about the theory. Theorem 1's assumptions A6 and A7 are too close to the success condition. A6 is essentially 'no false candidate passes the two conditions before the true lag,' which is what Phase 2 is supposed to establish. A7 uses Phase-1 selected lags, so it depends on CEDAR's own screening output rather than being a clean structural property of the true graph. The paper's own counterexample shows A1–A6 and A8 can hold while A7 fails, but that does not resolve the circularity. I would not cite Theorem 1 as a meaningful guarantee; the method should be judged on the finite-sample experiments, which are actually its strength.\n\nMinor issues: the stress-test table (Table 9) reports only 0.6 false negatives under 'A7 bypass×5,' which is hard to square with the theory's claim that a single A7-violating bypass on a true edge should cause a miss. Either the injected bypasses are not true A7 violations for the edges tested, or the theory is not predictive in finite samples. That deserves a sentence of clarification. Also, the claimed CAUSAL-TS package is not linkable from the text, and several tables report means without standard deviations. None of these are fatal.\n\nWho should read it: anyone working on low-sample lagged causal discovery in climate, hydrology, or finance, where T is small and AR(1)-like self-dynamics are plausible. It is not a framework paper; it is a method paper with solid empirical support in its intended regime.\n\nMy recommendation: send it to peer review. The empirical contribution is worth referee time, and the theoretical concerns can be addressed by tightening the assumptions, adding variance reporting, and making the code link available. A serious reviewer should push on A6/A7 but should not desk-reject.","headline":"A genuinely practical data-scarce causal-discovery method whose empirical claims are well supported, but whose oracle theorem rests on assumptions close to the desired success condition—worth reviewing seriously, not for the theory.","tokens_in":773,"tokens_out":819,"would_cite":true,"duration_ms":22083,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"CEDAR recovers lagged causal edges from short, sparse time series by using a deliberately small conditioning set and at most two independence tests per candidate edge.","keywords":["causal discovery","time series","lagged edges","conditional independence","distance correlation","autoregressive","MCI pruning","nonstationarity"],"falsifier":"Simulate three AR(1) processes with edges A(t-2)->C(t), A(t-2)->B(t), and B(t-1)->C(t), plus self-loops. Run CEDAR and check whether A(t-2)->C(t) is recovered. The paper predicts it should be missed because the path A(t-3)->B(t-1)->C(t) evades the conditioning set; if it is recovered, the claimed failure mode does not actually occur.","tokens_in":19893,"feed_emoji":"🔗","tokens_out":3649,"duration_ms":34905,"temperature":0.7,"pith_summary":"The paper proposes CEDAR, a constraint-based method for discovering which past values of one variable directly cause another in autoregressive time series. Its central claim is that in the low-sample regime, keeping conditioning sets small—by exploiting temporal structure rather than searching over large parent sets—preserves statistical power while remaining interpretable. The paper proves that, under a specific set of assumptions, CEDAR exactly recovers the eligible cross-variable lagged graph. Empirically, CEDAR is the best or tied-best method at T<=200 across three graph topologies up to d=75, and after regime segmentation it recovers 10 of 11 true edges in a real river network. The method's own appendix shows that the key path-blocking assumption can fail when lag-mismatched indirect paths exist, leading to missed edges.","feed_headline":"Two targeted tests recover lagged causes from short time series","feed_subtitle":"When T is small, small conditioning sets beat rich ones; trend confounds can be removed by a deterministic time node.","key_machinery":"The engine is the two-condition test. For a candidate edge X_c(t-w)->X_e(t), CEDAR first demands dependence given a conditioning set built from other candidates one step before their own lags; then demands independence of the deeper source X_c(t-w-1) from the target given the same set plus X_c(t-w). U-centered distance correlation on AR(1)-residualized targets screens lags nonparametrically, and a stable MCI pruning step rechecks every accepted edge on the same frozen graph, removing indirect links order-independently.","core_discovery":"CEDAR establishes that lagged cross-variable causation can be tested with at most two conditional-independence checks per candidate edge, accepting at most one lag per ordered pair. The first check keeps the candidate; the second verifies that conditioning on the selected lag blocks the source's deeper past. Under assumptions of Markovian sufficiency, faithfulness, strict AR(1) self-dynamics, a single true lag per pair, oracle candidate reachability, a path-blocking condition, and eligible-parent closure, the paper proves that the discovered graph equals the true eligible graph. The empirical claim is that this bounded conditioning delivers the best F1 among compared methods when T is betwee","pith_inferences":["The two-condition test suggests a general recipe: any marginal-association screening plus two targeted CI tests could be adapted to other data types, including discrete or event-based series.","The one-lag-per-pair restriction could be relaxed by iteratively residualizing accepted lags and re-running screening, potentially recovering multiple true lags per pair.","The path-blocking failure mode offers a diagnostic: when the deeper source remains conditionally dependent after the second test, that pair is precisely where a larger conditioning set is needed.","Applying changepoint detection before discovery, as done on the river data, is a transferable recipe for any regime-switching system."],"forward_implications":["Causal discovery from sparse, short time series becomes practical at higher dimensionality (d up to 75) when only a few lags survive screening.","Every reported edge is backed by explicit conditional-independence tests, making the output interpretable as evidence rather than a black-box score.","Deterministic trend-like nonstationarity can be absorbed by a time-indexed node, reducing spurious edges in real-world data.","The failure mode from lag-mismatched paths is bounded: CEDAR degrades smoothly, with F1 between 0.865 and 0.924 even under simultaneous assumption violations.","Segmenting a nonstationary system into regimes before discovery can expose clean causal signal, as shown by the river-network example."],"fun_headline_variants":["Short time series? Two CI tests reveal lagged causes","Bounded conditioning: Two tests suffice for lagged causality","CEDAR: Efficient lagged edge discovery for scarce data","At most two tests per edge recovers causal lags","When T is small, two tests beat richer conditioning"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"For every true edge, the source variable's deeper past must be fully blocked by the small conditioning set; the paper's own counterexample shows that a lag-mismatched path can slip through this set and cause the true edge to be missed.","fun_headline_variants_meta":{"raw":{"variants":["Short time series? Two CI tests reveal lagged causes","Bounded conditioning: Two tests suffice for lagged causality","CEDAR: Efficient lagged edge discovery for scarce data","At most two tests per edge recovers causal lags","When T is small, two tests beat richer conditioning"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000333,"raw_usage":{"total_tokens":1655,"prompt_tokens":677,"completion_tokens":978,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":421,"completion_tokens_details":{"reasoning_tokens":908}},"tokens_in":421,"tokens_out":978,"duration_ms":7076,"temperature":1.0,"reasoning_tokens":908,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T09:38:35.905941+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate three AR(1) processes with edges A(t-2)->C(t), A(t-2)->B(t), and B(t-1)->C(t), plus self-loops. Run CEDAR and check whether A(t-2)->C(t) is recovered. The paper predicts it should be missed because the path A(t-3)->B(t-1)->C(t) evades the conditioning set; if it is recovered, the claimed failure mode does not actually occur.","supporting_citations":[],"review_version":1}