Partitioning Error Components for Accuracy-Assessment of Near-Neighbor Methods of Imputation

By A. R. Stage, N. Crookston. Forestry sciences, 2007

56 citations recorded by the source corpus.

Imputation is applied for two quite different purposes: to supply missing data to complete a data set for subsequent modeling analyses or to estimate subpopulation totals. Error properties of the imputed values have different effects in these two contexts. We partition errors of imputation derived from similar observation units as arising from three sources: observation error, the distribution of observation units with respect to their similarity, and pure error given a particular choice of variables known for all observation units. Two new statistics based on this partitioning measure the accuracy of the imputations, facilitating comparison of imputation to alternative methods of estimation such as regression and comparison of alternative methods of imputation generally. Knowing the relative magnitude of the errors arising from these partitions can also guide efficient investment in obtaining additional data. We illustrate this partitioning using three extensive data sets from western North America. Application of this partitioning to compare near-neighbor imputation is illustrated for Mahalanobis- and two canonical correlation-based measures of similarity.

Subfields: Economics and Econometrics, Statistics and Probability

View Paper: https://www.semanticscholar.org/paper/648659478c69acdba67269618cc009865e6441b1

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Economics and Econometrics

In Statistics and Probability

Citations in the corpus

Nearest neighbor imputation of species-level, plot-scale forest structure attributes from LiDAR data (2008, methodology): “…the component of imputation error, apart frommeasurement error, that describes the variability in response values with the same predictor values around the true mean of the response values (Stage & Crookston, 2007), which is an important consideration for comparing the utility of imputation models.”
Spatial prediction of forest stand variables (2009, methodology): “Stage and Crookston (2007) developed closed-form expressions for approximations of the nearest neighbour prediction variance, with the variance being approximatively partitioned into several components.”
Evaluation of most similar neighbour and random forest methods for imputing forest inventory variables using data from target and auxiliary stands (2012, methodology): “SEI was suggested by Stage and Crookston (2007), who developed closed-form expressions for approximations of the nearest-neighbour prediction variance.”
yaImpute: An R Package for kNN Imputation (2008, methodology): “Because imputated values include diﬀerent components of error than those estimated using regression, Stage and Crookston (2007) exchanged the word error with diﬀerence in deﬁning root mean square diﬀerence ( rmsd ).”
Comparison of nearest neighbour approaches for small area estimation of tree species-specific forest inventory attributes in central Europe using airborne laser scanner data (2010, background): “In NN approaches, it appears convenient to name data with observed response variables reference observations and data for which only independent variables are available, target observations (Stage and Crookston 2007).”