Table 4.

Partial least squares regression (PLS) and simple linear regression (SLR) model performance metrics for Typha dry mass predictions.

PLS Typha Dry Mass Training Data Statistics
Response Variable:	sqrt [Typha Dry Mass]
Predictor Variables:	log[Ramet Area at 30 cm] + Inflorescence Presence + sqrt[Organic Matter Depth] + Total Ramet Height
PLS Components:	4
Cross-Validation Segments:	10
RMSECV Typha Dry Mass:	0.47 g
Explained Variance:	85.01%
SLR Typha Dry Mass Training Data Statistics
Response Variable:	sqrt [Typha Dry Mass)
Predictor Variable:	Total Ramet Height
Cross-Validation Segments:	10
RMSECV Typha Dry Mass:	2.27 g
Explained Variance:	18.38%
Typha Dry Mass Test Data Statistics
Number of Permutations:	1,000
Training/Test Replication:	n = 63 / n = 12
PLS DIFF (mean ± 1 sd):	-0.53 g ± 8.70 g
SLR DIFF (mean ± 1 sd):	-2.37 g ± 18.0 g

Training data statistics for PLS and SLR models present predictor variable(s), components selection (PLS only), k-fold cross-validation segments, root mean square error of cross-validation (RMSECV), and explained model variance. Test data statistics are given for permutation models results to estimate accuracy of external data predictions. Permutation results are given by mean ± 1 standard deviation of: DIFF = [predicted Typha dry mass – reference Typha dry mass]. All presented Typha dry mass results were back-transformed to show original data collection scale.