- Rating
- 1
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 1
- Soundness
- 1
Read review
Paper summary
The paper investigates causal structure identification and causal discovery using non-i.i.d. data. In particular, the authors consider a linear structural equation model in which both the edge coefficients and the noise variances are allowed to vary as functions of an additional variable. They first present a two-node example demonstrating that both residual mutual information and the score used by the DirectLiNGAM algorithm can incorrectly favor a causal direction opposite to the true data-generating process. The authors then propose an identification approach that is based on the curvature of the relationship between the additional variable and the corresponding conditional precision matrix.
Critical strengths
The paper addresses the sensitivity of existing methods to violations of constant coefficient and constant noise assumptions. To me, this is an interesting problem in causal discovery and I am not aware of work directly addressing it in the linear case.
Critical weaknesses
The presentation of the paper makes it very difficult to understand and assess the contribution. In its current form, it does not clearly communicate its findings to the reader. In the following, I state some examples from the first pages.
-
It is not clearly articulated that the paper is about linear structural equation models.
-
The central notion of coefficient modulation is introduced through the functional form , without a clear conceptual explanation of the "modulation" phenomenon, its practical relevance, or why this particular parameterization is appropriate.
-
Theorem 1 is a special example that shows that edge modulation can reverse the sign of the entropy. But the fact that this is an specific example needs to be inferred by the reader and is nowhere explicitly explained. The reader also has to infer that and . Moreover, the parameter is not defined properly and terms like "equiprobable" remain unexplained.
-
The last paragraph in Section 2 starts with the sentence "The packaged estimator placed first in constant-coefficient samples and moving-coefficient samples at each ". After reading multiple times I could infer that the numbers and refer to the number of simulated datasets in which the correct causal ordering was recovered when running DirectLiNGAM on 2 variables with sample sizes without and with edge modulation. However, the exact data generating process is not described, and, this could have been stated much more clearly for me.
-
Random variables are mixed upper and lower case.
-
At the bottom of the first page the authors suppose that "coefficients and relative noise variances are log-affine in the observed state". There is no explanation on this assumption. Is it a strong assumption? When can it be justified? Is it necessary?
-
Section 3 lacks clear explanations of intuition and there are no concrete examples.
Questions for authors
None.
2 comments
I thank the authors for their response and for addressing the specific examples I raised. However, my broader concern remains: in its current form, the manuscript does not communicate its findings clearly enough to the reader. The examples I gave are just some of the instances where the presentation could be improved. As I said I consider the problem interesting, but I believe that substantial improvements to the presentation are needed before the paper is ready. I therefore maintain my recommendation and encourage the authors to revise the presentation for a future conference submission.
RE: Official Review
Dear reviewer, thank you for your thoughtful comments and detailed review. We have revised the main text around the linear SEM, the matched reversal example, and the identification argument. Below we address your seven points in order.
In regard to the model class being unclear, the first sentence now states that the paper concerns linear structural equation models. Section 1 immediately gives Xi=∑j∈pa(i)Bij(V)Xj+εi. Conditional on the exogenous state V, the equations are linear. The DAG stays fixed; coefficients and noise variances may change with the state.
In regard to modulation appearing first as an unexplained formula, we now define it in words as a change in an edge's strength with the state, before introducing Bij(v)=βijeκijv. A one-unit increase in state multiplies the coefficient by eκij; κij=0 gives a constant effect. This form preserves the sign and gives an affine log magnitude. It is a sufficient shape restriction, rather than a claim that real systems generally obey an exponential law.
In regard to theorem 1's example and parameters being implicit, section 2 now introduces the construction explicitly: V,Ux,Uy independently take −1 and +1, each with probability 1/2; Zx,Zy are independent standard Gaussians; and X=2(V+Ux)/2Zx,Y=2mVX+2(V+Uy)/2Zy. Thus A=C=log2, b0=1, and κ=mlog2. The scalar m≥0 controls modulation. At m=0 the coefficient is 1 in both states; at m=1 it is 2 or 1/2. The noise process is identical. The theorem is now an existence result for this family, rather than a statement about every modulated model. It also certifies intervals, with the proof in Appendix AK.1.
In regard to the simulation sentence was hard to interpret, we now directly state the data-generating process, the number of datasets, and what counts as success. On 200 independent datasets of size n=2000 from the displayed family, DirectLiNGAM puts the true cause X first in 100% at m=0, 72% at .5, 20% at .6, and 0% at the tested values m≥.8. Table 40 reports the sample-size comparison. These finite-sample results are separate from the population certificate.
In regard to the overall notation being inconsistent, we have gone through and resolved these inconsistencies. Section 1 defines V as a random state and v as a value; X is the random vector; t indexes observations. The main theory uses B(v), Q(v), Σ(v) and Ω(v) consistently. Lower-case symbols in experimental appendices refer to generated or observed series.
In regard to the log-affine assumption lacking clear interpretation, section 3 now explains its role through the reverse-implied coefficient. With unrestricted coefficient functions, both directions can reproduce every conditional covariance. Log-affinity rules out the reverse coefficient away from a specific balance condition. Appendix AK.2 extends the sink criterion to an anchorwise convexity condition, while retaining log-affine relative noise variances. Section 6 shows failures outside that sufficient class without claiming universal failure there.
In regard to the identification argument lacking intuition and examples, section 3 now starts with the linear-prediction interpretation of the precision diagonal and a chain X1→X2→X3. At three states its signed curvatures are (.115,.115,0), identifying sink X3. After removing it and recomputing the marginal precision, they are (.115,0), recovering the order. A common curved noise factor shifts every gap equally to (−.285,−.285,−.400), showing why the minimum signed gap, rather than its absolute value, identifies the sink. Figure 1(b) illustrates the example.
We hope this makes the technical contribution substantially easier to assess. If any concern about the soundness of the results remains after these changes, we would be grateful to know which, so that we can address it directly.