<b>lmSubsets</b>: Exact Variable-Subset Selection in Linear Regression for <i>R</i>

Hofmann, M.; Gatu, Cristian; Kontoghiorghes, Erricos John; Colubi, Ana; Zeileis, Achim

doi:10.18637/jss.v093.i03

Cited by 13 publications

(7 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…All analyses were conducted in R version 3.5.3 (Free Software Foundation). Best subset selection was performed using the function lmSubsets() in the package lmSubsets ( 61 ). Given the exploratory nature of these analyses, we prespecified in our statisical analysis plan that we would not account for multiple comparisons when reporting P values or CIs.…”

Section: Methodsmentioning

confidence: 99%

Biomarkers of environmental enteric dysfunction are not consistently associated with linear growth velocity in rural Zimbabwean infants

Mutasa

Ntozini

Mbuya

et al. 2021

The American Journal of Clinical Nutrition

View full text Add to dashboard Cite

Background Child stunting remains a poorly understood, prevalent public health problem. Environmental enteric dysfunction (EED) is hypothesized to be an important underlying cause. Objectives Within a subgroup of 1169 children enrolled in the SHINE (Sanitation Hygiene Infant Nutrition Efficacy) trial in rural Zimbabwe, followed longitudinally from birth to 18 mo of age, we evaluated associations between the concentration of 11 EED biomarkers and linear growth velocity. Methods At infant ages 1, 3, 6, 12, and 18 mo, nurses measured child length and collected stool and blood; the lactulose-mannitol urine test was also conducted at all visits except at 1 mo. Stool neopterin, α-1 antitrypsin, myeloperoxidase, and regenerating gene 1β protein; urinary lactulose and mannitol; and plasma kynurenine, tryptophan, C-reactive protein, insulin-like growth factor-1 (IGF-1), soluble CD14, intestinal fatty acid binding protein, and citrulline were measured. We analyzed the change in relative [∆ length-for-age z score (LAZ)/mo] and absolute (∆ length/mo) growth velocity during 4 age intervals (1–3 mo; 3–6 mo; 6–12 mo; and 12–18 mo) per SD increase in biomarker concentration at the start of each age interval. Results In fully adjusted models, we observed only 3 small, statistically significant associations: kynurenine:tryptophan ratio at 12 mo was associated with decreased mean LAZ velocity during the 12–18 mo interval (−0.015 LAZ/mo; 95% CI: −0.029, −0.001 LAZ/mo); mannitol excretion at 6 mo was associated with increased LAZ velocity during the 6–12 mo interval (0.013 LAZ/mo; 95% CI: 0.001, 0.025 LAZ/mo), and plasma IGF-1 at 1 mo was associated with increased LAZ velocity during the 1–3 mo interval (0.118 LAZ/mo; 95% CI: 0.024, 0.211 LAZ/mo). Results for absolute growth velocity were similar, except IGF-1 was also associated with growth during the 12–18 mo interval. We found no other associations between any EED biomarker and linear growth velocity. Conclusions None of 11 biomarkers of EED were consistently associated with linear growth among Zimbabwean children. This trial was registered at clinicaltrials.gov as NCT01824940.

show abstract

Section: Methodsmentioning

confidence: 99%

Biomarkers of environmental enteric dysfunction are not consistently associated with linear growth velocity in rural Zimbabwean infants

Mutasa

Ntozini

Mbuya

et al. 2021

The American Journal of Clinical Nutrition

View full text Add to dashboard Cite

show abstract

“…To speed up, Furnival and Wilson (1974) introduced a well-known branch-and-bound algorithm with an efficient updating strategy for LMs, which was later implemented by R packages such as the leaps (Lumley and Miller 2017) and the bestglm (McLeod and Xu 2010). Hofmann, Gatu, Kontoghiorghes, Colubi, and Zeileis (2020) proposed an exact variable-subset selection algorithm for linear regression and implemented it in R package lmSubsets. Yet for GLMs, a simple exhaustive screen is undertaken in bestglm.…”

Section: Introductionmentioning

confidence: 99%

BeSS: An R Package for Best Subset Selection in Linear, Logistic and Cox Proportional Hazards Models

Wen¹,

Zhang²,

Quan³

et al. 2020

J. Stat. Soft.

View full text Add to dashboard Cite

We introduce a new R package, BeSS, for solving the best subset selection problem in linear, logistic and Cox's proportional hazard (CoxPH) models. It utilizes a highly efficient active set algorithm based on primal and dual variables, and supports sequential and golden search strategies for best subset selection. We provide a C++ implementation of the algorithm using an Rcpp interface. We demonstrate through numerical experiments based on enormous simulation and real datasets that the new BeSS package has competitive performance compared to other R packages for best subset selection purposes.

show abstract

“…In doing so, it also allows the identification of samples that do not have comparable estimates by competing models. The all-possible-subsets regression analysis was implemented using lmSubsets package [74] for R [75]. Table 2 lists the variables of the model that returned the smallest AIC value per GLCM and window size.…”

Section: Discussionmentioning

confidence: 99%

Estimation of woody plant species diversity during a dry season in a savanna environment using the spectral and textural information derived from WorldView-2 imagery

et al. 2020

View full text Add to dashboard Cite

Remote sensing techniques are useful in the monitoring of woody plant species diversity in different environments including in savanna vegetation types. However, the performance of satellite imagery in assessing woody plant species diversity in dry seasons has been understudied. This study aimed to assess the performance of multiple Gray Level Co-occurrence Matrices (GLCM) derived from individual bands of WorldView-2 satellite imagery to quantify woody plant species diversity in a savanna environment during the dry season. Woody plant species were counted in 220 plots (20 m radius) and subsequently converted to a continuous scale of the Shannon species diversity index. The index regressed against the GLCMs using the all-possible-subsets regression approach that builds competing models to choose from. Entropy GLCM yielded the best overall accuracy (adjusted R 2 : 0.41−0.46; Root Mean Square Error (RMSE): 0.60−0.58) in estimating species diversity. The effect of the number of predicting bands on species diversity estimation was also explored. Accuracy generally increased when three-five bands were used in models but stabilised or gradually decreased as more than five bands were used. Despite the peak accuracies achieved with three-five bands, performances still fared well for models that used fewer bands, showing the relevance of few bands for species diversity estimation. We also assessed the effect of GLCM window size (3×3, 5×5 and 7×7) on species diversity estimation and generally found inconsistent conclusions. These findings demonstrate the capability of GLCMs combined with high spatial resolution imagery in estimating woody plants species diversity in a savanna environment during the dry period. It is important to test the performance of species diversity estimation of similar environmental setups using widely available moderate-resolution imagery.

show abstract

lmSubsets: Exact Variable-Subset Selection in Linear Regression for R

Cited by 13 publications

References 21 publications

Biomarkers of environmental enteric dysfunction are not consistently associated with linear growth velocity in rural Zimbabwean infants

Biomarkers of environmental enteric dysfunction are not consistently associated with linear growth velocity in rural Zimbabwean infants

BeSS: An R Package for Best Subset Selection in Linear, Logistic and Cox Proportional Hazards Models

Estimation of woody plant species diversity during a dry season in a savanna environment using the spectral and textural information derived from WorldView-2 imagery

Contact Info

Product

Resources

About