Figures
Following the publication of this article [1] concerns were raised about the statistical analysis presented in this article. The journal has assessed these concerns with the help of a member of the Editorial Board and a statistical adviser, and concluded that although there were some issues with the reporting of the statistics, the overall results and conclusions of the article are supported. In order to update the reporting of the statistical analysis of this study, the following paragraphs are added:
The following paragraph is added to the Discussion section after paragraph 10:
The Kaiser criterion has limitations and therefore, the data was re-evaluated to determine the optimal number of dimensions to retain by applying Horn’s Parallel Analysis [2]. In addition, three other methods were used to confirm comprehensiveness or not of the results of the article. The three methods employed were the Marchenko-Pastur limit [3], the Gavish-Donoho method [4], and the Elbow method [5]. All these methods indicated retaining only one dimension, except for the Elbow method, which indicated retaining two dimensions. Therefore, as most statistical methods to determine the optimal number of dimensions to be retained coincided with the number of dimensions that were previously reported in the article, the results were unchanged.
The following paragraph is added to the subheading Internal consistency of the Results section after paragraph 1:
Cronbach’s α coefficient is considered one of the most popular methods for checking internal consistency. However, one of its limitations is the tau-equivalent. Therefore, to confirm the internal consistency of USAPS, the data was re-evaluated by applying the McDonald’s omega coefficient (ω) [6]. Because we assumed that USAPS was unidimensional, we used McDonald’s total omega coefficient [7]. Internal consistency with all items together was α = 0.81 and ω = 0.85, when excluding interaction α = 0.73 and ω = 0.80, excluding locomotion α = 0.74 and ω = 0.81, excluding head position α = 0.77 and ω = 0.83, excluding posture α = 0.80 and ω = 0.86, excluding activity α = 0.70 and ω = 0.81, and excluding appetite α = 0.88 and ω = 0.87. The McDonald’s omega coefficient results were close to those found by the Cronbach’s alpha coefficient originally used in the study and the classification of the interpretation actually improved from ’acceptable’ (α) to ’strong’ (ω) [7]. These findings confirm the robustness of the USAPS internal consistency.
The following paragraph is added to the subheading Criterion validity of the Results section after paragraph 1:
By a qualitative visual judgment, the dispersion of the relationship between USAPS and Numeric Scale is apparently non-monotonic. To comprehensively check the criterion validity result, the data was re-evaluated by applying four different methods to infer interdependencies between the USAPS and the Numeric Scale. First, Pearson’s correlation coefficient (rP) was applied, as suggested by Streiner et al. [8], to compare with the results given by Spearman’s rank correlation coefficient (rS). The correlation between USAPS and Numeric Scale showed values of rS = 0.83 and rP = 0.84. Therefore Pearson´s correlation did not change the interpretation of the Spearman´s one. Second, the contingency table built between USAPS and Numeric Scale was submitted to a chi-square test (χ2). This analysis evidenced the existence of a relationship between the variables (χ2 = 1741.6; df = 108; p< 2.2−16), according to S7 Table.
Third and fourth, a linear regression model and a quadratic regression model, respectively, applying USAPS as the predictor variable and Numeric Scale as the predictive variable were conducted. The quadratic modelling proved to be more adjusted than the linear one by the significance found in the analysis of variance (F = 65,206; p = 1,352−15) and other parameters described in S8 Table. The beta of the Numeric Scale of the quadratic regression was significant (p < 2.2−16), confirming the existence of a quadratic relationship between the variables. In light of this new analysis Fig 5 and its caption are updated. Please see the correct Fig 5 and Fig 5 caption below, assuming a quadratic relationship between both variables.
Scatter plot of the relationship between USAPS (predictor variable) and predictive variable Numeric Scale (A) with quadratic adjustment, and Simple Descriptive Scale (B), Visual Analogue Scale (C), and Facial Expression Scale (D) with linear adjustment. The line represents the fit line, and the shaded area is the standard error.
In Table 3, the links to the videos in the fourth column require updating. Please see the correct Table 3 with updated links here.
There is an error in the title for Table 4, “Load values, eigenvalues and variance of the USAPS items based on principal components analysis”. The title should read “Loading values, eigenvalues and variance of the USAPS items based on principal components analysis”. Please find the table with the correct title here.
USAPS–Unesp-Botucatu sheep acute composite pain scale. The structure was determined considering items with a loading value ≥ 0.50 or ≤ -0.50 (in bold), with representative dimension (eigenvalue > 1 and variance > 20%) [47].
Supporting information
S7 Table. Contingency table between USAPS (0–12) and Numeric Scale (1–10) submitted to chi-square test.
https://doi.org/10.1371/journal.pone.0268305.s001
(DOCX)
S8 Table. Model findings with linear and quadratic fit of the predictor variable USAPS and with the predictive variable Numeric Scale.
https://doi.org/10.1371/journal.pone.0268305.s002
(DOCX)
References
- 1. Silva NEOF, Trindade PHE, Oliveira AR, Taffarel MO, Moreira MAP, et al. Validation of the Unesp-Botucatu composite scale to assess acute postoperative abdominal pain in sheep (USAPS). PLoS One. 2020; 15(10): e0239622. https://doi.org/10.1371/journal.pone.0239622 pmid:33052903
- 2. Preacher KJ, MacCallum RC. Repairing Tom Swift’s Electric Factor Analysis Machine, Understanding Statistics. 2003; 2:1, 13–43. https://doi.org/10.1207/S15328031US0201_02
- 3. Dobriban E, Owen AB. Deterministic parallel analysis: an improved method for selecting factors and principal components. J R Stat Soc Series B. 2019; 81: 163–183. https://doi.org/10.1111/rssb.12301
- 4. Gavish M, Donoho DL. The optimal hard threshold for singular values is 4/√3, in IEEE Transactions on Information Theory. 2014; 60(8): 5040–5053. https://doi.org/10.1109/TIT.2014.2323359
- 5. Alkhayrat M, Aljnidi M, Aljoumaa KA. Comparative dimensionality reduction study in telecom customer segmentation using deep learning and PCA. J Big Data. 2020; 7:9. https://doi.org/10.1186/s40537-020-0286-0
- 6. McNeish D. Thanks coefficient alpha, we’ll take it from here. Psychol Methods. 2018; 23(3): 412–433. https://doi.org/10.1037/met0000144 pmid:28557467
- 7. Kalkbrenner MT. Alpha, omega, and H internal consistency reliability estimates: reviewing these options and when to use them. CORE, 1–12. https://doi.org/10.1080/21501378.2021.1940118
- 8.
Streiner D, Norman G, Cairney J. Health measurement scales: A practical guide to their development and use. 5th ed. New York: Oxford University Press Inc; 2015. https://doi.org/10.1093/acprof:oso/9780199231881.003.0006
Citation: Silva NEOF, Trindade PHE, Oliveira AR, Taffarel MO, Moreira MAP, Denadai R, et al. (2022) Correction: Validation of the Unesp-Botucatu composite scale to assess acute postoperative abdominal pain in sheep (USAPS). PLoS ONE 17(5): e0268305. https://doi.org/10.1371/journal.pone.0268305
Published: May 5, 2022
Copyright: © 2022 Silva et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.