A sequential distance-based approach for imputing missing data: Forward Imputation

Solaro, Nadia; Barbiero, Alessandro; Manzi, Giancarlo; Ferrari, Pier Alda

doi:10.1007/s11634-016-0243-0

Cited by 9 publications

(10 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We resort to the Page's test (Page 1963), which is based on punctual comparison of the coefficient empirical cumulative distribution functions. Following Solaro et al (2017), let F (M DC) , F (r) and F (r S ) be the cumulative distribution functions of the M DC, r and r s values in each iteration. We test the null hypothesis…”

Section: Simulation Results Under Non-normalitymentioning

confidence: 99%

A dependence measure flow tree through Monte Carlo simulations

Raffinetti

Ferrari

2020

Qual Quant

Self Cite

View full text Add to dashboard Cite

In applied psychological, behavioral and sociological research the majority of data are typically mixed (continuous and discrete) or, if continuous, they violate the normality condition. Given a dependent and an independent variables: a) both the variables may appear with distinct values (continuous variables); b) the dependent variable may present distinct values (continuous variable) and the independent variable tied values (discrete variable); c) the dependent variable may present tied values (discrete variable) and the independent variable distinct values (continuous variable). The dependence relationship between the variables could be assessed through the common correlation coefficients, i.e., the Pearson's, Spearman's and Kendall's coefficients, jointly with a recently revisited monotonic dependence coefficient, called "Monotonic Dependence Coefficient". But, the choice of the most suitable dependence measure in different scenarios may become problematic. The aim of the paper is to show which dependence measure to use to discover dependence relationships. A flow tree displaying how to find the best dependence measures is proposed by means of a Monte Carlo simulation study. Both Normal and non-Normal distributions producing continuous and discrete data, together with the possibility of transforming discrete data into continuous ones, are considered. Finally, validation of simulation findings on real data is also introduced.

show abstract

Section: Simulation Results Under Non-normalitymentioning

confidence: 99%

A dependence measure flow tree through Monte Carlo simulations

Raffinetti

Ferrari

2020

Qual Quant

Self Cite

View full text Add to dashboard Cite

show abstract

“…We focus on nonparametric methods that carry out single imputation of missing data according to different theoretical grounds. The first method is Forward Imputation (ForImp) [11,12], considered here in the two variants developed for quantitative data, i.e. ForImp with the Mahalanobis distance (ForImpMahalanobis -FIM in short) and ForImp with the Principal Component Analysis (ForImpPCA -FIP in short) [12].…”

Section: Imputing Quantitative Missing Data: the Considered Methodsmentioning

confidence: 99%

“…Motivated by several practical problems concerning missing data handling that we faced in a pure nonparametric perspective, we carried out an extensive investigation via simulation for inspecting and comparing the performance of three different nonparametric methods for single imputation of missing data, i.e. the Forward Imputation (ForImp) [11,12], the Iterative Principal Component Analysis method (IPCA) [13][14][15] and Stekhoven and Bühlmann's missForest method [16]. ForImp is a sequential, distance-based, distribution-free imputation procedure that is based on the nearestneighbour imputation method and can exploit a multivariate data analysis technique to synthesise the information of the complete part of the data [11,12].…”

Section: Introductionmentioning

confidence: 99%

“…the Forward Imputation (ForImp) [11,12], the Iterative Principal Component Analysis method (IPCA) [13][14][15] and Stekhoven and Bühlmann's missForest method [16]. ForImp is a sequential, distance-based, distribution-free imputation procedure that is based on the nearestneighbour imputation method and can exploit a multivariate data analysis technique to synthesise the information of the complete part of the data [11,12]. The IPCA method is an algorithmic-type technique that imputes missing values simultaneously by the iterative use of principal component analysis, recently rearranged by Josse et al [15] in the more general context of multiple imputation with factorial methods.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A simulation comparison of imputation methods for quantitative data in the presence of multiple data patterns

Solaro

Barbiero

Manzi

et al. 2018

Journal of Statistical Computation and Simulation

View full text Add to dashboard Cite

An extensive investigation via simulation is carried out with the aim of comparing three nonparametric, single imputation methods in the presence of multiple data patterns. The ultimate goal is to provide useful hints for users needing to quickly pick the most effective imputation method among the following: Forward Imputation (ForImp), considered in the two variants of ForImp with the Principal Component Analysis (PCA), which alternates the use of PCA and the Nearest-Neighbour Imputation (NNI) method in a forward, sequential procedure, and ForImp with the Mahalanobis distance, which involves the use of the Mahalanobis distance when performing NNI; the iterative PCA technique, which imputes missing values simultaneously via PCA; the missForest method, which is based on random forests and is developed for mixed-type data. Performance of these methods is compared under several data patterns characterized by different levels of kurtosis or skewness and correlation structures.

show abstract

“…(2) Forward Imputation (ForImp) [18], a sequential distance-based approach based on the nearest-neighbour imputation method and available in the R package GenForImp [17], with its two variants for quantitative data: ForImp with PCA (FIP) and ForImp with the Mahalanobis distance (FIM). In order to involve the clinical grouping variable into the imputation process, we have introduced a third variant: Within-Group FIP (WG.FIP), that is, FIP run within each clinical group, rather than on the whole dataset.…”

Section: Imputation Methodsmentioning

confidence: 99%

Handling Missing Data in Observational Clinical Studies Concerning Cardiovascular Risk: An Insight into Critical Aspects

2017

Self Cite

View full text Add to dashboard Cite

In observational clinical studies, subjects' health status is empirically assessed according to research protocols that prescribe aspects to investigate and methods for investigation. Commonly to many fields of research, these studies are frequently affected by incompleteness of information, a problem that, if not duly handled, may seriously invalidate conclusions drawn from investigations. Regarding cardiovascular risk assessment, coronary risk factors (e.g. high blood pressure) and proxies of neurovegetative domain (e.g. heart rate variability) are individually evaluated through direct measurements taken in laboratory. A major cause of missingness can be ascribed to the fact that overall sets of collected data typically derive from aggregation of a multitude of sub-studies, undertaken at different times and under slightly different protocols that might not involve the same variables. Data on certain variables can thus be missing if such variables were not included in all protocols. This issue is addressed by referring to a clinical case study concerning the role of Autonomic Nervous System in the evaluation of subjects' health status.

show abstract

A sequential distance-based approach for imputing missing data: Forward Imputation

Cited by 9 publications

References 17 publications

A dependence measure flow tree through Monte Carlo simulations

A dependence measure flow tree through Monte Carlo simulations

A simulation comparison of imputation methods for quantitative data in the presence of multiple data patterns

Handling Missing Data in Observational Clinical Studies Concerning Cardiovascular Risk: An Insight into Critical Aspects

Contact Info

Product

Resources

About