Average Standard Error Of Prediction Research Articles

Quantitative structure–property relationship study of standard formation enthalpies of acyclic alkanes using atom-type-based AI topological indices

A quantitative structure–property relationship (QSPR) study was performed for prediction of enthalpies of 134 acyclic alkanes using modified Xu (mXu) index and atom-type-based AI topological indices. At first, a simple linear regression model was developed using mXu index alone and the statistics were R2=0.947, F=2335 and standard error of 1.00. The results showed that combination of the atom-type-based AI topological indices and mXu index can produce significant improvement in the statistical quality of the model, especially the decrease in the standard error was 33% relative to the simple linear model. The final model was validated to be statistically significant and reliable using external validation technique. External validation was performed by dividing the entire data set into three subsets and predicting enthalpy values for each subset from the other two as training sets. Average standard error of calibration of 0.66 and average standard error of prediction of 0.68 demonstrated the validity and good efficiency of the topological indices in modeling enthalpies of alkanes. The obtained results showed that the enthalpy for acyclic alkanes is dominated by molecular size and the atomic groups are also important although their contributions are much smaller than that of the molecular size.

Wheat Moisture Measurement with a Fringing Field Capacitive Sensor

Grain storage managers could improve the quality of stored grain if they could directly monitor stored grain moisture content, which is a key indicator of stored grain quality and an early indicator of deterioration. However, shortcomings of currently available sensors have prevented them from achieving widespread acceptance in the industry. A new fringing field capacitive (FFC) sensor was tested to determine its suitability and accuracy for moisture content measurements in grain. Sensors were calibrated at temperature from 10°C to 30°C using six samples of hard red winter (HRW) wheat from three locations and two crop years. The polynomial calibration models had standard error of prediction (SEP) values that averaged 0.68% wet basis (w.b.) moisture content for data not corrected for bulk density. The average SEP improved to 0.50% w.b. when the readings were corrected based on sample bulk density, yielding a 95% confidence interval of ±1.0% w.b. for these data. The measured sensor accuracy, close to that of laboratory instruments, is appropriate for an in situ instrument for monitoring stored grain and for rapid determination of grain moisture content in bulk containers.

Heuristic extraction of rules in pruned artificial neural networks models used for quantifying highly overlapping chromatographic peaks.

The suitability of an approach for extracting heuristic rules from trained artificial neural networks (ANNs) pruned by a regularization method and with architectures designed by evolutionary computation for quantifying highly overlapping chromatographic peaks is demonstrated. The ANN input data are estimated by the Levenberg-Marquardt method in the form of a four-parameter Weibull curve associated with the profile of the chromatographic band. To test this approach, two N-methylcarbamate pesticides, carbofuran and propoxur, were quantified using a classic peroxyoxalate chemiluminescence reaction as a detection system for chromatographic analysis. Straightforward network topologies (one and two outputs models) allow the analytes to be quantified in concentration ratios ranging from 1:7 to 5:1 with an average standard error of prediction for the generalization test of 2.7 and 2.3% for carbofuran and propoxur, respectively. The reduced dimensions of the selected ANN architectures, especially those obtained after using heuristic rules, allowed simple quantification equations to be developed that transform the input variables into output variables. These equations can be easily interpreted from a chemical point of view to attain quantitative analytical information regarding the effect of both analytes on the characteristics of chromatographic bands, namely profile, dispersion, peak height, and residence time.

Quantitative Determination of Moisture in Lubricants by Fourier Transform Infrared Spectroscopy

This paper describes the development of a practical Fourier transform infrared (FT-IR) method for the determination of moisture in lubricants through the combined use of signal transduction and differential spectroscopy to circumvent matrix effects. The acid-catalyzed stoichiometric reaction of 2,2-dimethoxypropane (DMP) with moisture to produce acetone was used to provide IR signals proportional to the amount of moisture present in oils. Calibration standards were prepared by spiking polyalphaolefin (PAO) gravimetrically with water using dioxane as a carrier. For FT-IR analysis, standards and samples were diluted with acidified isooctane and then split, with one aliquot treated with DMP and the other with a blank reagent. The spectra of the two aliquots were collected, and a differential spectrum was obtained so as to ratio out the invariant spectral contributions from the sample. Quantitation for moisture was based on measurement of the peak height of the nu(C=O) absorption of acetone at 1717 cm(-1), yielding a standard error of calibration of approximately 40 ppm H2O. The method was validated by standard addition of water in dioxane to PAO containing added base as well as to new and used oils. In all cases the method responded quantitatively to standard addition, the average standard error of prediction being approximately 80 ppm, with the results showing only a minor dependence on the oil formulation. From an analytical perspective, the FT-IR method is both more reproducible and more accurate than Karl Fischer methods and has advantages in terms of environmental considerations, sample size, and speed of analysis as well as the variety of oil types that can be handled. Signal transduction/differential spectroscopy may have broader utility as an alternative means for the determination of low levels of moisture in complex matrices.

Beyond average protein secondary structure content prediction using FTIR spectroscopy

This paper demonstrates that secondary structure information beyond purely protein secondary structure content can be predicted from FTIR (Fourier transform infrared spectroscopy) spectra of proteins with a high degree of accuracy. Both neural networks and adaptive neuro-fuzzy inference systems (ANFISs) were employed to predict helix/sheet segment information. The best results were achieved using ANFISs with fuzzy subtractive clustering based on normalised, compressed amide I data with an average SEP (standard error of prediction, root mean of squared errors) of 1.51. Predictions for average helix/sheet length based merely on the amide I band maximum position in combination with the full-width at half-height resulted in a comparable average SEP of 1.62. This suggests the importance of information on the position and width of the amide I band maximum for the prediction of helix/sheet segment information. Finally, the most promising pattern recognition approaches found in this study were applied to a protein with an as yet unknown x-ray structure: native a1-antichymotrypsin (a1-ACT).

Empirical modeling of summer lake surface temperatures in southwest Greenland

This work presents a method to estimate mean daily lake surface water temperatures using only air temperature, theoretical clear‐sky solar radiation, and lake size. Surface water temperatures were measured at a selection of lakes in southwest Greenland during the summers of 1998–2000. The lakes are small (surface area <150 ha) with maximum depths ranging from 3.5 to 47 m. An empirical model requiring only local air temperature and theoretical clear‐sky solar radiation is developed to predict daily mean lake surface temperatures in summer for each lake. The model approximates the slow integrated response of water temperature to meteorological forcing by applying an exponential smoothing filter to air temperature. Exponential smoothing results in a 35% improvement in model fit compared with a model using unsmoothed air temperatures. The smoothed air temperatures and clear‐sky solar radiation are linearly combined to estimate the daily mean lake surface temperatures. The smoothing parameters and the three linear coefficients of the model, obtained individually for each of 15 lakes, are found to relate to lake area and maximum depth, leading to the development of a general model. With this general model it is possible to predict the summer surface temperatures at any lake in this region where local air temperatures can be estimated. Cross‐validation of the general model at each lake in turn indicated a 90% forecast skill and average standard error of prediction of 1.0°C. Examination of the daily prediction errors over time suggests a relation to strong wind events.

Robust on‐line measurement of conversion and molecular weight using NIR spectroscopy during solution polymerization

AbstractFiber‐optic near‐infrared (NIR) spectroscopy was used to monitor the monomer conversion and the weight‐average molecular weight of the polymer produced during solution polymerization of methyl methacrylate (MMA) carried out in a lab‐scale reactor. NIR spectra were recorded during batch and semi‐continuous reactions using an in situ transmission probe. Off‐line gravimetry and GPC were used as reference methods to provide the conversion and the average molecular weight data set required for the calibration procedure. A statistical model was generated using partial least‐squares regression (PLS) to relate the NIR spectral data to the two polymerization variables of interest. The measurements were then validated for various operating conditions (i.e., different solvent, initiator, MMA, and chain‐transfer agent concentrations) and for both batch and semi‐continuous modes. The conversion was predicted during three validation experiments with an average standard error of prediction (SEP) of 2.1%. The on‐line evaluation of M̄w was obtained with an average relative SEP of 5.7%; such on‐line NIR measurement was thus demonstrated to be robust and accurate, even in the case of versatile use of the polymerization plant. © 2002 Wiley Periodicals, Inc. J Appl Polym Sci 85: 2510–2520, 2002

Design of angle-tolerant multivariate optical elements for chemical imaging.

Multivariate optical elements (MOEs) are multilayer optical interference coatings with arbitrary spectral profiles that are used in multivariate pattern recognition to perform the task of projecting magnitudes of special basis functions (regression vectors) out of optical spectra. Because MOEs depend on optical interference effects, their performance is sensitive to the angle of incidence of incident light. This angle dependence complicates their use in imaging applications. We report a method for the design of angle-insensitive MOEs based on modification of a previously described nonlinear optimization algorithm. This algorithm operates when the effects of deviant angles of incidence are simulated prior to optimization, which treats the angular deviation as an interferent in the measurement. To demonstrate the algorithm, a 13-layer imaging MOE (IMOE, with alternating layers of high-index Nb2O5 and low-index SiO2) for the determination of Bismarck Brown dye in mixtures of Bismarck Brown and Crystal Violet, was designed and its performance simulated. For angles of incidence that range from 42 degrees to 48 degrees, the IMOE has an average standard error of prediction (SEP) of 0.55 microM for Bismarck Brown. This compares with a SEP of 2.8 microM for a MOE designed by a fixed-angle algorithm.

Non-Invasive Fermentation Analysis Using an Artificial Neural Network Algorithm for Processing near Infrared Spectra

The feasibility of using Artificial Neural Networks (ANN) to improve the performance of a fibre optic, near infrared (NIR) spectroscopic probe for monitoring fermentation processes was investigated. A miniature diode array spectrometer, operating between 1100 and 1450 nm, was used for on-line, in situ fermentation monitoring by placing a bifurcated fibre bundle inside a fermentation vessel. Non-linearities in the spectral response were complicated by the combined effects of spectral variations in the OH vibrational bands due to temperature fluctuations and pH variation. As the fermentation proceeds interferences from other absorbing molecules and cell masses increase at an exponential rate. The feasibility of accurately predicting both glucose and ethanol concentrations simultaneously during a fermentation process were assessed by applying Partial Least Squares (PLS) and Artificial Neural Networks (ANN) to the NIR spectral data. For a 5% glucose fermentation, a PLS model was able to predict “on-line” concentrations with standard errors of prediction ( SEP) of 0.19% for glucose and 0.11% for ethanol. Three separate on-line fermentation experiments were performed in order to determine the possibility of using a model developed from one experiment to predict the concentrations of another experiment. PLS models produced an average SEP of 0.21% when used to predict different fermentation experiments. The ANN algorithm produced an average SEP of 0.13% on the same data.

Partial Least-Squares Modeling of Near-Infrared Reflectance Data for Noninvasive in Vivo Determination of Deep-Tissue pH

Noninvasive monitoring of deep-tissue pH has been demonstrated with the use of near-infrared spectroscopic measurements and the partial least-squares (PLS) multivariate calibration technique. The near-infrared reflectance spectra (700 to 1100 nm) of the teres major muscle in five New Zealand rabbits were obtained in vivo, along with reference pH values in the muscle measured by microelectrodes. The muscle pH was varied by controlling the blood supply to the muscle. PLS analysis with cross-validation techniques, along with several data preprocessing methods, was used to relate the tissue pH to spectra. When multi-subject PLS calibration models were used to predict a new independent subject, a subject-dependent offset was observed. Several strategies for minimizing the subject-dependent offset were discussed. With a baseline subtraction procedure, the subject-dependent offset was minimized to less than 0.1 pH units while the average standard error of prediction (SEP) was close to 0.05 pH units. This result suggests that it is possible to build a single robust calibration model for all new independent subjects. Tissue chemistry during ischemia (blood flow reduction) is different from the chemistry of reperfusion (blood flow restoration), and it was found that separate calibration models permit more accurate prediction of pH.

Low-Volume Diffuse-Transmittance Fiber-Optic Near-IR Spectrophotometer

A simple, inexpensive near-IR spectrophotometer is described that employs single fiber-optic strands in a diffuse-transmission geometry. The instrument can examine unusually small sample volumes and offers sufficient throughput to perform multicomponent analyses on diffusely transmitting powdered samples. Control of the effective optical pathlength is found to be critical in obtaining reproducible results when scattering samples are to be analyzed. A pseudo-double-beam arrangement is employed to reduce instrument noise and to set the optical pathlength independently for each sample. Sample sizes of only a few tens of milligrams are required for analysis. The resulting performance is nearly equivalent to that of a commercial reflectance instrument. An average standard error of prediction of 2.77% w/w was obtained for four-component mixtures of benzoic acid derivatives. The single-fiber transmission geometry is shown to be useful also in the near-IR analysis of microliter volumes of liquids. A standard error of cross-validation of 17 mM sodium chloride dissolved in water was achieved.

ALTERNATIVE PROGENY TESTING PROGRAMS FOR WEANING WEIGHT AND EASE OF CALVING IN BEEF CATTLE

Progeny test programs were characterized by numbers of young bulls, successful matings (conceptions) per young bull, herds in which each young sire was used, proven bulls used as reference sires and herd size. Matings to reference sires were calculated as a percentage of the number of young sire matings. Designs were simulated and mixed-model equations formed and added to equations obtained from the Canadian ROP program for years 1971–1978. The average standard error of prediction (SEP) of young bulls was used for comparing designs. Increasing matings per young sire decreased SEP curvilinearly. The curvilinearity was greater when either herd size or herds per young sire was low. Increasing herd size and herds per young sire improved connectedness between sires and reduced SEP. Number of young sires affected SEP only when fewer than 10 bulls were sampled. Number of young sire matings were more important in reducing SEP than percent reference sire matings. If number of young sire matings was adequate to achieve a desired level of SEP, 10% reference sire matings were sufficient. The number of reference sires used did not affect SEP. Design patterns affected SEP of weaning weight (h2 = 0.25) and calving ease (h2 = 0.10) similarly. SEP was not affected by the number of proven sires in a breed (Maine Anjou, 133 sires or Blonde d’Aquitaine, 39 sires) or by the connectedness between proven sires.

Predicting True Metabolizable Energy, Gross Energy, Carbohydrate, and Proximate Analysis Values by Assuming Additivity

Seven feedingstuffs and 42 mixtures thereof were assayed under routine quality control conditions for true metabolizable energy (TME), gross energy, proximate constituents, and starch plus sugars expressed as glucose. Comparisons were made between the observed analytical values of the mixtures and the values predicted from the data describing their constituent parts by assuming additivity.The largest differences between observed and predicted values were associated with a small number of mixtures and may have resulted from errors in mixing, changes in moisture content, or analytical variation. The differences between observed and predicted TME values were similar to the differences in the other analytical values; the average standard error of prediction was about 2.5% of the observed TME value.