Variable selection in large environmental data sets using principal components analysis

Abstract

In many large environmental datasets redundant variables can be discarded without the loss of extra variation. Principal components analysis can be used to select those variables that contain the most information. Using an environmental dataset consisting of 36 meteorological variables spanning 37 years, four methods of variable selection are examined along with dierent criteria levels for deciding on the number of variables to retain. Procrustes analysis, a measure of similarity and bivariate plots are used to assess the success of the alternative variable selection methods and criteria levels in extracting representative variables. The Broken-stick model is a consistent approach to choosing signi®cant principal components and is chosen here as the more suitable criterion in combination with a selection method that requires one principal component analysis and retains variables by starting with selection from the ®rst component.

Differing conceptual maps of skills for implementing evidence-based interventions held by community-based organization practitioners and academics: A multidimensional scaling comparison

“…the two maps showed a pattern of similarities that were not random and possessed a comparable underlying structure). Based on a previous categorization scheme for m 2 values, 0.63 suggests average similarity, although caution in applying the categories is warranted since they are not validated within a sample of GCM studies [ 43 ]. These findings suggest that the two maps were more similar than expected due to chance alone, yet there were meaningful levels of dissimilarity.…”

Turbidity‐discharge hysteresis in a meso‐scale catchment: The importance of intermediate scale events

“…In the PCA (Figure 7) the first two axes explain 48% of the variation. According to King and Jackson (1999) parameter selection method, weather related parameters, seasons and hysteresis patterns were all important. Clockwise events were associated with winter and spring season and Q parameters, that is, maximum HFQ, maximum snowmelt (maxSM).…”

Defining dual-axis landscape gradients of human influence for studying ecological processes

“…When each city was considered independently, the same three axes explained between 42.60 and 54.89% of the variance (Table 1 in S1 File ), demonstrating the scalability of emergent landscape gradients across scales. However, using the broken stick method [ 51 ], only the first two axes exceeded the 22.1% cumulative variance threshold for combined and city-specific analyses. The principal component explaining the largest proportion of data variation for the combined data (16.7%) was strongly negative for developed land-cover classes, with neutral or positive loadings for forested, open, and agricultural classes ( Table 3 ).…”