Journal of research of the National Bureau of Standards (1977)最新文献

Herein we study the problem of assessing, on the basis of noisy and incomplete observations, how much information there is in the data for model identification in compartmental systems. The underlying concept is that of an "information distance" between competing models, and estimation of this distance on the basis of the given data is discussed. Useful reduction of the dimensionality of the corresponding least squares problem is accomplished by regarding the decay rate constants as primary parameters of interest and the other parameters of the model as nuisance parameters. Estimation of the decay rate function is also discussed.

Chromatographic fingerprinting of complex biological samples is an active research area with a large and growing literature. Multivariate statistical and pattern recognition techniques can be effective methods for the analyisis of such complex data. However, the classification of complex samples on the basis of their chromatographic profiles is complicated by two factors: 1) confounding of the desired group information by experimental variables or other systematic variations, and 2) random or chance classification effects with linear discriminants. We will treat several current projects involving these effects and methods for dealing with the effects. Complex chromatographic data sets often contain information dependent on experimental variables as well as information which differentiates between classes. The existence of these types of complicating relationships is an innate part of fingerprint-type data. ADAPT, an interactive computer software system, has the clustering, mapping, and statistical tools necessary to identify and study these effects in realistically large data sets. In one study, pattern recognition analysis of 144 pyrochromatograms (PyGCs) from cultured skin fibroblasts was used to differentiate cystic fibrosis carriers from presumed normal donors. Several experimental variables (donor gender, chromatographic column number, etc.) were involved in relationships that had to be separated from the sought relationships. Notwithstanding these effects, discriminants were developed from the chromatographic peaks that assigned a given PyGC to its respective class (CF carrier vs normal) largely on the basis of the desired pathological difference. In another study, gas chromatographic profiles of cuticular hydrocarbon extracts obtained from 179 fire ants were analyzed using pattern recognition methods to seek relations with social caste and colony. Confounding relationships were studied by logistic regression. The data analysis techniques used in these two example studies will be presented. Previously, Monte Carlo simulation studies were carried out to assess the probability of chance classification for nonparametric and parametric linear discriminants. The level of expected chance classification as a function of the number of observations, the dimensionality, and the class membership distributions were examined. These simulation studies established limits on the approaches that can be taken with real data sets so that chance classifications are improbable.

This paper presents a technique based on the intuitively-simple concepts of Sample Domain and Effective Prediction Domain, for dealing with linear regression situations involving collinearity of any degree of severity. The Effective Prediction Domain (EPD) clarifies the concept of collinearity, and leads to conclusions that are quantitative and practically useful. The method allows for the presence of expansion terms among the regressors, and requires no changes when dealing with such situations.

Consideration of certain aspects of scientific method leads to discussion of recent research on the role of screening designs in the improvement of quality. A projective rationale for the use of these designs in the circumstances of factor sparsity is advanced. In this circumstance the possibility of identification of sparse dispersion effects as well as sparse location effect is considered. A new method for the analysis of fractional factorial designs is advanced.

A consistent set of thermochemical property values, Δ _f H°, Δ _f G°, S°, and $C_p^o$ , at 298.15 K is given for the known constituents of aqueous sulfur dioxide ( ${\text{SO}}_2^0\left(\text{aq}\right)$ , ${\text{HSO}}_3^{-}\left(\text{aq}\right)$ , ${\text{SO}}_3^{2-}\left(\text{aq}\right)$ , H⁺(aq), and ${\text{S}}_2{\text{O}}_5^{2-}\left(\text{aq}\right)$ ). Also tabulated are values of the mean ionic activity coefficients, osmotic coefficients, partial pressure of SO₂(g), and the relative apparent molar enthalpy as a function of concentration of SO₂(aq) at 298.15 K. The data analysis considered a wide variety of measurement techniques: calorimetric enthalpies of solution and reaction, heat capacities, equilibrium constants, solubilities, and vapor pressure measurements, both partial and total, over aqueous solutions of SO₂ for the temperature range 278 to 393 K. All auxiliary data have been taken from the most recent set of CODATA values which were converted to a standard state pressure of one bar (0.1 MPa). For the process ${\text{SO}}_2\left(\text{g}\right)={\text{SO}}_2^0\left(\text{aq}\right)$ , the selected "best" values are: K = 1.23±0.05 mol kg^-1 bar^-1, ΔG° = -0.5±0.10 kJ mol^-1, ΔH° = -26.97±0.30 kJ mol^-1, and $\Delta C_p^o=155\pm 10\text{J}{\text{mol}}^{-1}{\text{K}}^{-1}$ . The standard state partial molar entropy of ${\text{SO}}_3^{2-}\left(\text{aq}\right)$ , obtained by the analysis of data via two independent thermodynamic pathways is -15.40±0.80 J mol^-1 K^-l at 298.15 K. Parameters are given which extend the predictions to temperatures up to 373 K.

The NBS-2 balance was designed and built at NBS and transferred to the BIPM in 1972. It is presently used for the comparison of national prototype kilograms with international standards. Excellent environmental conditions at the BIPM have resulted in a long-term standard deviation of 1 microgram (l× 10^-9) for a comparison of two 1-kilogram standards. With this remarkable precision, one has begun to observe and quantify systematic biases of less than 5 micrograms. The nature of these biases is presented as well as the remedy adopted to eliminate their influence on both the final measurement results and the variance assigned to those results.

Triple-point-of-succinonitrile cells have been tested and established as Standard Reference Material (SRM) 1970. Of the 115 cells tested, 109 were accepted as SRM 1970. Five of the 115 cells had triple-point temperatures lower than 58.0785 °C (the low-temperature limit established for SRM 1970) and, consequently, were rejected. One of the 115 cells broke during tests on it. The mean value of the triple-point temperatures (obtained by freezing) of the 109 cells is 58.0796±0.0015 °C, where the uncertainty is the total estimated uncertainty relative to the International Practical Temperature Scale of 1968, Amended Edition of 1975. The standard deviation of the triple-point temperatures is 0.48 mK. The purity of the succinonitrile of the SRM 1970 cells is estimated to range from 99.999,97% to 99.999,84%. The preparation of the cells, the various tests performed on them, and the procedure recommended for their use are described.

This paper discusses the factors known to influence the performance of automatic speech recognizers and describes test procedures for characterizing their performance. It is directed toward all the stakeholders in the speech community (researchers, vendors and users) consequently, the discussion of test procedures is not directed toward the needs of specific users to demonstrate the performance characteristics of any one specific algorithmic approach or particular product. It relies significantly on contributions from an emerging consensus standards activity, especially material developed within the IEEE Working Group on Speech I/O Performance Assessment.