Fig. 2

The ability of NIRS to classify Ae. albopictus mosquitoes as being young or old. a The best-fit regression coefficient function for a model trained on laboratory-reared mosquitoes showing the most informative regions of the spectrum. Grey lines show best-fit model for each of the 100 dataset randomisations whilst black line indicates average. b Ability of the model to predict age classification of laboratory-reared mosquitoes. Histogram of the estimated linear predictor for the test observations colour-coded by the true class (green, true young mosquitoes; blue, true old mosquitoes). Vertical black line indicates optimum threshold for classifying mosquitoes as old or young (“left” predicted to be young, “right” predicted to be old). The shaded area where two distributions overlap corresponds to misclassified test observations, false negatives to the left and false positives to the right of the optimal classification threshold. c The corresponding confusion matrix for the best model trained and predicting laboratory-reared mosquitoes showing the different error rates: tnr, true negative rate; fnr, false negative rate (specificity); fpr, false positive rate; and tpr, true positive rate (sensitivity). d The receiver operating characteristic (ROC) curve for the best-fit model predicting laboratory-reared mosquitoes showing the false positive and true positive rates achievable for different classification probability thresholds (shifting the black vertical line (b) left or right) whilst the overall performance is given by the area under the ROC curve (AUC). The pink dashed line denotes a model with no predictive ability (a random chance of correct prediction) whilst a perfect model with 100% sensitivity and specificity would be in the top left corner (coordinates 0, 1). The solid line shows the average ROC curve; boxplots show the variability for 100 randomisations of the training, validation and testing datasets (box edges, inner and outer whiskers show 25th/75th, 15th/85th and 5th/95th percentiles, respectively; black line inside the box showing the median/50th-percentile). e The ROC curve showing the ability of the model trained on laboratory rerared mosquitoes to predict the age classification of mosquitoes reared in the field-derived environment. f The corresponding confusion matrix of the best model