A Caution Regarding Rules of Thumb for Variance Inflation Factors

O’Brien, Robert M.

doi:10.1007/s11135-006-9018-6

Cited by 7,014 publications

(3,703 citation statements)

References 8 publications

Supporting

Mentioning

3,395

Contrasting

Unclassified

Order By: Relevance

“…Only two variables in one set of regressions exceed this threshold (SOM Table S5). It appears that previously suggested alternatives to multiple regression have statistical effects that may be worse than using a multiple regression with mutually correlated predictors (O'Brien, 2007).…”

Section: Analytical Frameworkmentioning

confidence: 99%

“…Regressions using either one or multiple independent variables were run. When multiple regressions were run, we also assessed multi-colinearity of the independent (explanatory) variables with Tolerance and Variance Inflation Factors (O'Brien, 2007). Caper was also used to run phylogenetically informed t-tests evaluating whether taxa with a consistently patent promontorial artery have a smaller ratio of stapedial canal area to promontorial canal area than taxa that frequently have an involuted promontorial artery.…”

Section: Statistical Analysesmentioning

confidence: 99%

“…However, slightly lower degrees of multicollinearity can apparently make the pattern of significance of multiple regression less tractable when predictor variables are removed or added. A typical rule of thumb is that levels of multicollinearity should have a tolerance below 0.10 (tolerance ¼ 1 e r 2 ) or a variance inflation factor (VIF) below 10 (VIF ¼ 1/tolerance) (O'Brien, 2007). Only two variables in one set of regressions exceed this threshold (SOM Table S5).…”

Section: Analytical Frameworkmentioning

confidence: 99%

See 2 more Smart Citations

Internal carotid arterial canal size and scaling in Euarchonta: Re-assessing implications for arterial patency and phylogenetic relationships in early fossil primates

Boyer

Kirk

Silcox

et al. 2016

Journal of Human Evolution

View full text Add to dashboard Cite

a b s t r a c tPrimate species typically differ from other mammals in having bony canals that enclose the branches of the internal carotid artery (ICA) as they pass through the middle ear. The presence and relative size of these canals varies among major primate clades. As a result, differences in the anatomy of the canals for the promontorial and stapedial branches of the ICA have been cited as evidence of either haplorhine or strepsirrhine affinities among otherwise enigmatic early fossil euprimates. Here we use micro X-ray computed tomography to compile the largest quantitative dataset on ICA canal sizes. The data suggest greater variation of the ICA canals within some groups than has been previously appreciated. For example, Lepilemur and Avahi differ from most other lemuriforms in having a larger promontorial canal than stapedial canal. Furthermore, various lemurids are intraspecifically variable in relative canal size, with the promontorial canal being larger than the stapedial canal in some individuals but not others. In species where the promontorial artery supplies the brain with blood, the size of the promontorial canal is significantly correlated with endocranial volume (ECV). Among species with alternate routes of encephalic blood supply, the promontorial canal is highly reduced relative to ECV, and correlated with both ECV and cranium size. Ancestral state reconstructions incorporating data from fossils suggest that the last common ancestor of living primates had promontorial and stapedial canals that were similar to each other in size and large relative to ECV. We conclude that the plesiomorphic condition for crown primates is to have a patent promontorial artery supplying the brain and a patent stapedial artery for various nonencephalic structures. This inferred ancestral condition is exhibited by treeshrews and most early fossil euprimates, while extant primates exhibit reduction in one canal or another. The only early fossils deviating from this plesiomorphic condition are Adapis parisiensis with a reduced promontorial canal, and Rooneyia and Mahgarita with reduced stapedial canals.

show abstract

Section: Analytical Frameworkmentioning

confidence: 99%

Section: Statistical Analysesmentioning

confidence: 99%

Section: Analytical Frameworkmentioning

confidence: 99%

See 1 more Smart Citation

Internal carotid arterial canal size and scaling in Euarchonta: Re-assessing implications for arterial patency and phylogenetic relationships in early fossil primates

Boyer

Kirk

Silcox

et al. 2016

Journal of Human Evolution

View full text Add to dashboard Cite

show abstract

“…Before conducting regression models, multicollinearity was checked using the variance inflation factor (VIF) for each explanatory variable. No problem of multicollinearity was detected, as the highest VIF value (all VIFs < 1.30) was well below the thresholds (≥ 5 or ≥ 10) generally considered as evidence of multicollinearity (see O’brien, 2007). …”

Section: Methodsmentioning

confidence: 95%

Non-medical use of prescription drugs by young men: impact of potentially traumatic events and of social-environmental stressors

Rougemont-Bücking

Grazioli

Marmet

et al. 2018

European Journal of Psychotraumatology

View full text Add to dashboard Cite

Background: Non-medical use of prescription drugs (NMUPD) is an increasing phenomenon associated with physical and psychological consequences. This study investigated the effects of distinct forms of stress on NMUPD.Methods: Data from 5308 young adult men from the Swiss cohort study on substance use risk factors (C-SURF) were analysed regarding NMUPD of sleeping pills, tranquilizers, opioid analgesics, psychostimulants, and antidepressants. Various forms of stress (discrete, potentially traumatic events, recent and long-lasting social-environmental stressors) during the period preceding the NMUPD assessment were measured. Backward log-binomial regression was performed and risk ratios (RR) were calculated.Results: NMUPD was significantly associated with the cumulative number of potentially traumatic events (e.g. for opioid analgesics, RR = 1.92, p < .001), with problems within the family (e.g. for sleeping pills, RR = 2.45, p < .001), and the peer group (e.g. for tranquilizer use, RR = 2.34, p < .01). Factors describing family functioning in childhood showed very few significant associations. Sexual assault by acquaintances was associated only with use of sleeping pills (RR = 2.91, p p <.01); physical assault by acquaintances was not associated with NMUPD. Physical (e.g. for psychostimulants, RR = 2.01, p < .001) or sexual assaults (e.g. for antidepressants, RR = 4.64, p < .001) perpetrated outside the family context did show associations with several drug categories.Conclusion: NMUPD appears to be more consistently associated with discrete and potentially traumatic events and with recent social-environmental stressors than with long-lasting stressors due to family functioning during childhood and youth. Physical and sexual assaults perpetrated by strangers showed more associations with NMUPD than those perpetrated by a family member.

show abstract

“…A tolerance of less than 0.20 or 0.10 and/or a VIF of 5 or 10 and above were shown to indicate a multicollinearity problem [2]. In the statistical program we use (SPSS), VIF and tolerance values are used to detect multicollinearity in linear regression analysis; however, in the logistic regression analysis the "correlation of estimates" test automatically removes those…”

mentioning

confidence: 99%

Reply to the letter to the editor concerning: ‘Relationship between increased carotid artery stiffness and idiopathic subjective tinnitus’

Bayraktar

Taşolar

2017

Eur Arch Otorhinolaryngol

View full text Add to dashboard Cite

from the analysis. When we perform a correlation analysis (Spearman or Kendall tau) between the relevant variables, the coefficient of r is lower than 0.7 for all variables. According to these results, there is no multicollinearity in our study; for this reason we did not need to perform any regression modelling to find the association between the variables with moderate to high degree of tinnitus.Second, we stated in the method section that the systolic and diastolic blood pressures were recorded after 5 min of rest. Since our study design was cross-sectional, we performed these measurements before the Doppler ultrasonographic examinations to exclude the individual fluctuations for the blood pressure. Even though we have not specified this deficiency in the text, we have done these measurements at least three times from each arm and have received the highest average of three measurements. So, we did not think any regression dilution bias to our manuscript. To the Editor,We thank the authors for their valuable evaluation of our article entitled "Relationship between increased carotid artery stiffness and idiopathic subjective tinnitus" published in European Archives of OtoRhino-Laryngology in 2017 [1]. First of all, the authors are right about the details of multiple regression analysis, but we also conducted both steps as they mentioned. Since the lack of the issue that the variables for unadjusted p values <0.10 were identified in multivariate logistic regression analysis is not specified in the statistical analysis section, we thank you for your attention in this matter.The Variance Inflation Factor (VIF) and tolerance are both widely used measures of the degree of multicollinearity of the independent variable with the other independent variables in regression models. A tolerance of less than 0.20 or 0.10 and/or a VIF of 5 or 10 and above were shown to indicate a multicollinearity problem [2]. In the statistical program we use (SPSS), VIF and tolerance values are used to detect multicollinearity in linear regression analysis; however, in the logistic regression analysis the "correlation of estimates" test automatically removes those This reply refers to the comment available at

show abstract

A Caution Regarding Rules of Thumb for Variance Inflation Factors

Abstract: multi-collinearity, tolerance, variance inflation factors, variance of regression coefficients,

Cited by 7,014 publications

References 8 publications

Internal carotid arterial canal size and scaling in Euarchonta: Re-assessing implications for arterial patency and phylogenetic relationships in early fossil primates

Internal carotid arterial canal size and scaling in Euarchonta: Re-assessing implications for arterial patency and phylogenetic relationships in early fossil primates

Non-medical use of prescription drugs by young men: impact of potentially traumatic events and of social-environmental stressors

Reply to the letter to the editor concerning: ‘Relationship between increased carotid artery stiffness and idiopathic subjective tinnitus’

Contact Info

Product

Resources

About