Cross-validated Root Mean Square Research Articles

Prediction of Oil and Oleic Acid Concentrations in Individual Corn (Zea Mays L.) Kernels Using Near-Infrared Reflectance Hyperspectral Imaging and Multivariate Analysis

Due to their heterogeneous structure and variability in form, individual corn (Zea mays L.) kernels present an optical challenge for nondestructive spectroscopic determination of their chemical composition. Increasing demand in agricultural science for knowledge of specific traits in kernels is driving the need to find high-throughput methods of examination. In this study macroscopic near-infrared (NIR) reflectance hyperspectral imaging was used to measure small sets of kernels in the spectroscopic range of 950 nm to 1700 nm. Image analysis and principal component analysis (PCA) were used to determine kernel germ from endosperm regions as well as to define individual kernels as objects out of sets of kernels. Partial least squares (PLS) analysis was used to predict oil or oleic acid concentrations derived from germ or full kernel spectra. The relative precision of the minimum cross-validated root mean square error (RMSECV) and root mean square error of prediction (RMSEP) for oil and oleic acid concentration were compared for two sets of two hundred kernels. An optimal statistical prediction method was determined using a limited set of wavelengths selected by a genetic algorithm. Given these parameters, oil content was predicted with an RMSEP of 0.7% and oleic acid content with an RMSEP of 14% for a given corn kernel.

Estimating average daytime and daily temperature profiles within Europe

We present a methodology for estimating the average profiles of daytime and daily ambient temperature from a spatially-continuous database for any location within Europe. The primary database with 1-km grid resolution was developed by interpolation of monthly averages of 7 daily values of temperature: minimum and maximum and 5 measurements at 3-h intervals from 6:00 to 18:00 hours Greenwich Mean Time. With a little over 800 meteorological stations available, we obtained a cross-validation root mean square error of 1.0–1.2 °C, while the interpolation error is lower, at 0.5–0.7 °C. A polynomial fit was applied to estimate the daytime temperature profile (assuming only time from sunrise to sunset) from the interpolated 3-h measurements for each month. The curve fit coefficients make it possible to calculate a number of derived data, such as average daytime temperature, maximum daytime temperature and time of its occurrence within the region. An example demonstrates the coupling of the simulated daytime temperature profile with a model for assessing the relative efficiency of electricity generation by crystalline silicon photovoltaic modules. As an alternative to the polynomial fitting, a double-cosine method was applied to enable calculation of daily (24-h) temperature profiles for each month using interpolated minimum and maximum temperatures. Compared to the polynomial curve-fitting, this method does not offer lower errors, but it provides data which are more suitable for estimation of solar thermal heating or calculation of degree days for building heating/cooling.

Effects of within-lake variability of fossil assemblages on quantitative chironomid-inferred temperature reconstruction

We studied the effects of within-lake variation of subfossil chironomid assemblages on chironomid-inferred temperatures in five lakes in southern Norway. In each lake six replicate surface-sediment samples in the deepest part of the lake basin and two transects of seven samples each towards the littoral were analysed for subfossil chironomid assemblages. A quantitative chironomid–July air temperature inference model applied to the subfossil assemblages inferred temperatures with a standard deviation of 0.35–0.54°C for the six mid-lake samples and of 0.38–0.59°C for all 20 sediment samples. This variation seems high when compared with the cross-validated root mean square error of prediction (RMSEP) of the model of 1.04°C. However, we show that the variability of the replicate mid-lake samples can only account for a maximum of 15% of the model prediction error if the RMSEP is decomposed into a component due to uncertainties associated with sampling the chironomid assemblage of a lake and into a residual error component. Thus, according to our results analysing multiple samples in the model lakes to obtain better estimates of their subfossil assemblage composition may only slightly reduce the model prediction error. It was difficult to detect a clear pattern of chironomid-inferred temperatures with respect to water depth in any single lake due to the high variability of inferred values. However, a comparison of all five study lakes revealed that, relative to the mid-lake samples, cooler temperatures were inferred at intermediate depths and warmer temperatures in the shallowest parts of the lakes, although two shallow-water samples with very low inferred temperatures were exceptions in this respect. This depth-related bias of inferred temperatures in our relatively shallow study lakes indicates that chironomid–temperature models calibrated on mid-lake samples should only be applied with caution to near-shore sediments and that a systematic offset in the inferred temperatures may occur.

Shortwave–near infrared spectroscopy for non-destructive determination ofmaturity of wine grapes

Near-infrared (NIR) spectroscopy has become a very popular technique for thenon-invasive assessment of intact fruit. This work presents an application ofa low-cost commercially available NIR spectrometer for the estimationof ripeness of Chilean wine grapes. Two configurations for the spectraacquisition were used (diffuse transmittance and interactance), using acustom-designed contact probe. Samples of Chardonnay, Carménère andCabernet Sauvignon, collected over the 2002 harvest and pre-harvest seasons,were analysed for total soluble solids content (°Brix).Partial least squares calibration models, obtained from several preprocessingtechniques (smoothing, multiplicative signal correction, standard normal variate,etc), were compared. Also, two spectral regions were used, one without the redpart of the visible spectrum (just the short-wave (SW–NIR) region) and the otherincluding it. Performance of different models was assessed in terms of root meansquare of cross-validation, root mean square of prediction (RMSEP) andR2for a validation set of samples. RMSEPs of 1.06 with R2 = 0.942indicate that it is possible to estimate wine grape ripeness (°Brixvalue), by using a CCD portable spectrometer. The red grape models performedbetter than the white grape models.

Regression spline models and model calibration in the identification of protein storage conditions

The study of protein-construct activity provides valuable insight in the early stages of drug discovery. Identification of optimal storage conditions that maintain activity of the constructs is an important issue. In the study reported herein, a space-filling design was used to assess the effects of eight design variables—buffer, pH, NaCl, protein concentration reducing agent detergent, MgCl2, and temperature—on protein activity. A regression spline analysis is presented and settings of the explanatory factors resulting in the best predicted protein activity over the explored space are identified. The models are selected initially based on the Akaike information criterion and later assessed via root mean square error and cross-validation root mean square error. Detergent, buffer, temperature, and smooth functions of pH and protein concentration were found to have large effects on protein activity. The models were calibrated via calculation of the empirical distribution of the cross-validation root mean square error. In this framework several models provide a similar fit, and further experimentation is required for more definitive conclusions re- garding protein storage conditions to be made. The cross-validation root mean square error calibration of models is recommended for applications involving comparisons of models with respect to their predictive ability.

Near-infrared mass median particle size determination of lactose monohydrate, evaluating several chemometric approaches.

The influence of particle size on near-infra red (NIR) spectra is typically considered a 'nuisance factor' which many scatter correction methods attempt to eliminate, e.g., multiplicative scatter correction. However, particle size is a key issue in the formulation of many pharmaceutical products and has a profound effect on the behaviour of both raw materials and drug substances during formulation. NIR has already been demonstrated as a potential alternative particle sizing technique to current accepted methodology. This investigation assessed several chemometric approaches that model this information, using lactose monohydrate as the raw material. A variety of modelling techniques were applied to both zero order and second derivative spectra namely multiple linear regression, partial least squares, principal component regression and artificial neural networks. One further data transformation evaluated was polar coordinates, although no statistical data were generated. Typically, cross-validation root mean square errors of calibration and cross-validation root mean square errors of prediction of approximately 5 microns were calculated for all of the modelling techniques. These values are comparable to those associated with the reference technique (laser diffractometry). Correlation coefficients of approximately 0.98 for all techniques were also calculated. The predictive abilities for models generated using second derivative spectra were found to be comparable to those obtained using zero order spectra.