Assessing the Impact of Energy Retrofits on Indoor Climate Conditions Using Mixed Effects Models: Methodology and R Implementation

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

1. External vapor pressure is identified as the dominant predictor of indoor humidity (β̂ ≈ 0.63) , but the physical pathways (infiltration, diffusion, ventilation) responsible for this coupling are not discussed.
2. The retrofit effect is found to increase indoor vapor pressure (β̂ ≈ 0.35) , yet the manuscript does not sufficiently explain whether this is due to reduced air exchange (airtightness), increased moisture retention or inadequate ventilation post-retrofit.
3. The study does not incorporate any first-principles modeling (e.g., moisture balance, diffusion, or advection equations). Even a simplified moisture balance framework would strengthen the manuscript by providing a physical basis for interpreting regression coefficients.
4. The use of stepwise removal of non-significant variables (p > 0.05) requires further justification.
5. Given the hourly time-series nature of the data, temporal autocorrelation is expected. Although the Durbin-Watson statistic is reported (≈1.87) , this alone may not be sufficient. The authors should discuss potential autocorrelation more thoroughly, and consider models incorporating temporal structure (e.g., AR(1)).

Author Response

Responses in the attached document

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The current manuscript presents a well-structured methodological framework for evaluating energy retrofit impacts on indoor climate conditions using linear mixed effects models, illustrated with a case study of indoor water vapor pressure in 23 Irish homes. After a detailed reading of the presented research, it can be concluded that the manuscript is well written and deals with actual topics. The reviewer recommends that the manuscript be accepted after the revision of the following issues：

1. Section 2.1.4, The authors state that missing data are handled under the Missing At Random (MAR) assumption using maximum likelihood. However, no justification is given for why MAR is plausible in this setting. Please discuss the sensitivity of results to violations of MAR and consider reporting the proportion of missingness per variable.
2. Section 2.3.1, Why did the authors use stepwise removal of non-significant fixed effects based on p>0.05? This approach is known to produce biased standard errors and inflated Type I error rates.
3. Fig. 2, The pre- and post-retrofit panels are described as “near-identical”, why present both? A single panel with a note that the relationship is unchanged would be sufficient.
4. Table 2, the summary statistics for external and room variables are shown only for the combined pre/post periods? The table heading says “grouped by retrofit status” but external variables have a single column “mean(SD)” – this appears to be overall, not by status. Please split by retrofit status or correct the caption.
5. Section 3.7, the holdout validation uses a 70/30 split within the same 23 homes. Why? This tests temporal generalizability, not generalizability to new homes.
6. Lines 589–592, the authors also acknowledge this limitation, but the term “external validation” is misleading. It is recommended that the authors should explicitly acknowledge this limitation and clarify that the validation supports temporal reliability within the sample rather than cross‑sample transferability.

Author Response

Responses in attached document

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This paper introduces a linear mixed-effects (LME) modeling framework for quantifying the impacts of building energy retrofits on indoor climate conditions, which demonstrates clear methodological innovation. Nevertheless, substantial revisions are necessary to enhance statistical rigor and mechanistic depth of interpretation:
1.    While the manuscript cites comprehensive reviews by Fisk et al. (2020) and Wang et al. (2022), it lacks a systematic overview of empirical applications of LME models in building performance and indoor environmental quality research. A targeted literature synthesis on this specific theme is strongly encouraged.
2.    The selection of water vapor partial pressure (Pw)—instead of relative humidity (RH)—as the primary response variable is physically grounded, yet insufficiently connected to widely accepted thermal comfort standards such as ASHRAE Standard 55. The authors should incorporate conversion relationships between Pw and core thermal comfort metrics (e.g., PMV/PPD) and provide quantitative discussion linking critical Pw thresholds to mold growth risks.
3.    The justification for the random-effects structure (1 + RetrofitStatus + ExternalPw | HomeID) is inadequate: no formal statistical tests for random slope selection are reported, and no explanation is given for omitting random slopes for room temperature (RoomT) or season. A detailed table presenting full likelihood ratio test (LRT) outcomes is required.
4.    The treatment of missing data lacks rigor. The assertion that “imputation was not necessary” is not validated by reported variable-specific missing proportions or diagnostic tests for missing data mechanisms (MCAR/MAR/MNAR). The authors should add Little’s MCAR test or graphical visualization of missing data patterns.
5.    Figure 2 reveals a strong negative correlation between random intercepts and ExternalPw slopes (ρ = −0.79), which is interpreted as “physically plausible”. However, the manuscript does not rule out mathematical artefacts arising from model specification, nor does it validate the robustness of this correlation across different climatic zones. Supplementary simulation analyses or cross regional validation are recommended.
6.    The mechanistic explanation of the retrofit effect remains superficial. The observed 0.09 kPa increase in post retrofit Pw is not linked to a quantitative decomposition of contributions from enhanced airtightness and reduced ventilation rates, nor are differential effects of distinct retrofit measures (e.g., external wall insulation versus window replacement) explored. Decomposition of building level heterogeneity via random effects is strongly advised.
7.    The translation of a 0.09 kPa Pw rise to a “4–5 percentage point increase in RH” is not supported by location specific calculations of annual hours with RH above 60% under Irish climatic conditions. The analysis does not integrate well established mold growth models (e.g., the Vereecken–Roels model) to quantify associated health risks. Computation of a formal mold growth index is highly recommended.
8.    The elevated humidity observed in other rooms is attributed exclusively to “proximity to moisture sources”, without investigating mediating effects of ventilation frequency or room functional usage. Further mechanistic analysis is needed to validate this interpretation.
9.    The conclusion that retrofits elevate indoor humidity lacks contextual justification specific to Ireland’s maritime climate, retrofit driven airtightness gains, and insufficient natural ventilation. The discussion of broader generalizability to other regions and building stocks remains underdeveloped.

Author Response

Responses in attached document.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

The revised manuscript has fully addressed all previous review comments. All modifications and explanations are reasonable and sufficient. No further revisions are required, and the paper is recommended for acceptance.