Stepwise deletion: a technique for missing-data handling in multivariate analysis

Hemel, Jan; Vandervoet, H; Hindriks, Fr; Vanderslik, W

doi:10.1016/s0003-2670(00)86157-7

Cited by 11 publications

(3 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Data imputation, however, will be inaccurate in cases where almost all observations of a given variable are missing. In this situation, it is sometimes better to exclude the variable from the analysis, particularly if it is highly correlated to other variables (Hemel, van der Voet, van der Hindriks, & Slik, 1987). This way, variable exclusion is less likely to impact analyses or at least will have lower impact, as seen for the smaller negative correlation of fully absent variables to PC estimation and PCA similarity (Table 3).…”

Section: Discussionmentioning

confidence: 99%

Schrödinger's phenotypes: Herbarium specimens show two‐dimensional images are both good and (not so) bad sources of morphological data

2020

View full text Add to dashboard Cite

1. Museum specimens are the main source of information on organisms' morphological features. Although access to this information was commonly limited to researchers able to visit collections, it is now becoming freely available thanks to the digitization of museum specimens. With these images, we will be able to collectively build large-scale morphological datasets, but these will only be useful if the limits to this approach are well-known. To establish these limits, we used two-dimensional images of plant specimens to test the precision and accuracy of image-based data and analyses. 2. To test measurement precision and accuracy, we compared leaf measurements taken from specimens and images of the same specimens. Then, we used legacy morphometric datasets to establish differences in the quality of datasets and multivariate analyses between specimens and images. To do so, we compared the multivariate space based on original legacy data to spaces built with datasets simulating image-based data. 3. We found that trait measurements made from images are as precise as those obtained directly from specimens, but as traits diminish in size, the accuracy drops as well. This decrease in accuracy, however, has a very low impact on dataset and analysis quality. The main problem with image-based datasets comes from missing observations due to image resolution or organ overlapping. Missing data lowers the accuracy of datasets and multivariate analyses. Although the effect is not strong, this decrease in accuracy suggests caution is needed when designing morphological research that will rely on digitized specimens. 4. As highlighted by images of plant specimens, 2D images are reliable measurement sources, even though resolution issues lower accuracy for small traits. At the same time, the impossibility of observing particular traits affects the quality of image-based datasets and, thus, of derived analyses. Despite these issues, gathering phenotypic data from two-dimensional images is valid and may support largescale studies on the morphology and evolution of a wide diversity of organisms.

show abstract

Section: Discussionmentioning

confidence: 99%

Schrödinger's phenotypes: Herbarium specimens show two‐dimensional images are both good and (not so) bad sources of morphological data

2020

View full text Add to dashboard Cite

show abstract

“…Data imputation, however, will be inaccurate in cases where almost all observations of a given variable are missing. In this situation, it is sometimes better to exclude the variables from the analysis, particularly if it is highly correlated to other variables (Hemel et al, 1987). This way, variable exclusion is less likely to impact analyses or at least will have lower impact, as seen for the smaller negative correlation of fully absent variables to PC estimation and PCA similarity (Table 3).…”

Section: Discussionmentioning

confidence: 99%

Schrödinger’s phenotypes: herbarium specimens show two-dimensional images are both good and (not so) bad sources of morphological data

Borges

Reis

Izbicki

2020

Preprint

View full text Add to dashboard Cite

AbstractMuseum specimens are the main source of information on organisms’ morphological features. Although access to this information was commonly limited to researchers able to visit collections, it is now becoming freely available thanks to the digitization of museum specimens. With these images, we will be able to collectively build large-scale morphological datasets, but these will only be useful if the limits to this approach are well-known. To establish these limits, we used two-dimensional images of plant specimens to test the precision and accuracy of image-based data and analyses.To test measurement precision and accuracy, we compared leaf measurements taken from specimens and images of the same specimens. Then we used legacy morphometric datasets to establish differences in the quality of datasets and multivariate analyses between specimens and images. To do so, we compared the multivariate space based on original legacy data to spaces built with datasets simulating image-based data.We found that trait measurements made from images are as precise as those obtained directly from specimens, but as traits diminish in size, the accuracy drops as well. This decrease in accuracy, however, has a very low impact on dataset and analysis quality. The main problem with image-based datasets comes from missing observations due to image resolution or organ overlapping. Missing data lowers the accuracy of datasets and multivariate analyses. Although the effect is not strong, this decrease in accuracy suggests caution is needed when designing morphological research that will rely on digitized specimens.As highlighted by images of plant specimens, 2D images are reliable measurement sources, even though resolution issues lower accuracy for small traits. At the same time, the impossibility of observing particular traits affects the quality of image-based datasets and, thus, of derived analyses. Despite these issues, gathering phenotypic data from two-dimensional images is valid and may support large-scale studies on the morphology and evolution of a wide diversity of organisms.

show abstract

“…Excluded ob.jccts came about equally from all classes. (The arbitrariness of this way of missing data removal has stimulated us in a later study to develop another missing data handling technique, finally emerging into the method of slepwise delelion [24]. )…”

Section: Preprocessing the Datamentioning

confidence: 99%

Exploring multivariate clinical chemical routine data concerning three major disease groups

Hemel

Vandervoet

Hendriks

et al. 1988

Journal of Analytical Methods in Chemistry

Self Cite

View full text Add to dashboard Cite

In preparation for multivariate analysis, an exploratory study has been undertaken to investigate the relative position, separability, homogeneity and shape of three major disease groups, using data from a clinical chemical routine package. The data set consists of 46 hepatology patients, 50 nephrology patients and 46 cardiology patients, and the measured blood levels include 20 common clinical chemical routine assays. Missing value problems were avoided by deleting some of the variables and objects. A univariate analysis was used as the basis ofa rescaling of the data. Bivariate (pairwise) plots of some major assays each show limited separation. The set of three such plots of the three major principal components reveals more distinction between the groups than was offered by univariate analysis. Three-dimensional extensions of these techniques allow better insight than any of the two-dimensional plots, but these three-dimensional versions require more plots for complete interpretation. Non-linear mapping of the data is the best way of retaining the distances and a fairly good separation is achieved in the plot. The plot is less informative about shape and relative position of the classes. Representation of the data as pictures of faces does not offer additional information and visual clustering is worse than in any of the techniques mentioned. During the analysis many assumed properties of the data are confirmed and a good starting pointfor multivariate classification is attained. Easy visual detection of outliers is offered by all techniques. Unfortunately, valuable information is lost in this data set by deleting some incomplete variables.

show abstract

Stepwise deletion: a technique for missing-data handling in multivariate analysis

Cited by 11 publications

References 6 publications

Schrödinger's phenotypes: Herbarium specimens show two‐dimensional images are both good and (not so) bad sources of morphological data

Schrödinger's phenotypes: Herbarium specimens show two‐dimensional images are both good and (not so) bad sources of morphological data

Schrödinger’s phenotypes: herbarium specimens show two-dimensional images are both good and (not so) bad sources of morphological data

Exploring multivariate clinical chemical routine data concerning three major disease groups

Contact Info

Product

Resources

About