Skip to main content
. 2020 Nov 23;13:591. doi: 10.1186/s13071-020-04468-6

Table 4.

Performance of Models 1 and 2 for Savitzky–Golay transformed spectra of faecal samples from validation dataset Val1, Val2 and Val3 for the prediction of haemoglobin in faeces collected from South Australia within the wavelength range of 387–609 nm

Model Validation dataseta
Val1 (n = 39) Val2 (n = 35) Val3 (n = 33)
r2p RMSEP (µg Hb/mg faeces) %SN (95% CI) %SP (95% CI) r2p RMSEP (µg Hb/mg faeces) %SN (95% CI) %SP (95% CI) r2p RMSEP (µg Hb/mg faeces) %SN (95% CI) %SP (95% CI)
1 0.05 2.78 71.4% (47.8–88.7%) 44.4% (21.5–69.2%) 0.73 1.57 88.2% (63.6–98.5%) 72.2% (46.5–90.3%) 0.77 1.67 88.9% (65.3–98.6%) 64.3% (35.1–87.2%)
2 NA 3.03 57.1% (34.0–78.2%) 66.7% (41.0–86.7%) 0.68 1.73 82.4% (56.6–96.2%) 77.8% (52.4–93.6%) 0.76 1.73 94.4% (72.7–99.9%) 78.6% (49.2–95.3%)

r2p, Coefficient of determination for correlation of prediction; RMSEP, root-mean-squared error of prediction (units in µg Hb/mg faeces); %SN (95% CI): sensitivity of the model with a 95% confidence interval; %SP (95% CI), specificity of the model with a 95% CI

aTo include variations that may be present in future validations, a randomly selected portion of faecal samples included in Models 1 and 2 were mixed with various concentrations of haemoglobin and scanned as a slab on separate days using the visual–near-infrared spectrometer (Val1 and Val2). Faecal samples were collected from the Lower North and Yorke Peninsula in South Australia to represent locations not included in the models as validation dataset 3 (Val3)