Graphical diagnostics for regression model determinations with consideration of the bias/variance trade-off

Estimates of model parameters (regression coefficients forming the regression vector) for a multivariate linear model have been the subject of considerable discussion. Regression diagnostics utilized in chemometrics for a multivariate linear model are often based on a single number such as the coefficient of determination, root mean square error of cross-validation, selectivity, etc. Additionally, regression diagnostics commonly applied focus on model bias and do not include variance or model complexity. This paper demonstrates that substantial information is available through a graphical study of trends in model parameters as determined by plots of regression diagnostics using bias, variance, and/or model complexity measures. Also illustrated is that by using harmonious graphics which simultaneously use bias and variance information, determination of proper model parameters without cross-validation is possible. This paper concludes with comments on the next level of regression diagnostics, including use of color, sound, and virtual reality.     

A novel approach to deriving closed form expressions for frequencystability variance measures

An approach is developed for obtaining closed-form expressions for variance estimators of frequency stability as functions of several parameters. It is based on a direct integration method and therefore expressions for autocorrelation functions of phase and frequency instabilities are not needed. The method is particularly useful when the transfer function approach of J. Rutman (1974) and E. Boileau (1978) is used, according to which the integral expressions of the variance estimate are chosen a priori. Using this method, the integrals can be evaluated without the need of calculating the corresponding sample of the variance. The relationship between this approach and the structure function characterization of frequency stability developed by W.C. Lindsey and C.M. Chie (1976) is also presented     

A sub‐1V supply CMOS voltage reference generator

An integrated sub‐1V voltage reference generator, designed in standard 90‐nm CMOS technology, is presented in this paper. The proposed voltage reference circuit consists of a conventional bandgap core based on the use of p‐n‐p substrate vertical bipolar devices and a voltage‐to‐current converter. The former produces a current with a positive temperature coefficient (TC), whereas the latter translates the emitter‐base voltage of the core p‐n‐p bipolar device to a current with a negative TC. The circuit includes two operational amplifiers with a rail‐to‐rail output stage for enabling stable and robust operation overall process and supply voltage variations while it employs a total resistance of less than 600 K Ω. Detailed analysis is presented to demonstrate that the proposed circuit technique enables die area reduction. The presented voltage reference generator exhibits a PSRR of 52.78 dB and a TC of 23.66ppm/^°C in the range of ? 40 and 125^°C at the typical corner case at 1 V. The output reference voltage of 510 mV achieves a total absolute variation of ± 3.3% overall process and supply voltage variations and a total standard deviation, σ, of 4.5 mV, respectively, in the temperature range of ? 36 and 125^°C. Copyright © 2011 John Wiley & Sons, Ltd.     

Assessing Disclosure Risk and Data Utility Trade-off in Transaction Data Anonymization

Organizations and businesses, including financial institutions and healthcare providers, are increasingly collecting and disseminating information about individuals in the form of transactions. A transaction associates an individual with a set of items, each representing a potentially confidential activity, such as the purchase of a stock or the diagnosis of a disease. Thus, transaction data need to be shared in a way that preserves individuals'' privacy, while remaining useful in intended tasks. While algorithms for anonymizing transaction data have been developed, the issue of how to achieve a "desired" balance between disclosure risk and data utility has not been investigated. In this paper, we assess the balance offered by popular algorithms using the R-U confidentiality map. Our analysis and experiments shed light on how the joint impact on disclosure risk and data utility can be examined, which allows the production of high-quality anonymization solutions.     

Wavelength selection for multivariate calibration using tikhonov regularization

Prediction of sample properties using spectroscopic data with multivariate calibration is often enhanced by wavelength selection. This paper reports on a built-in wavelength selection method in which the estimated regression vector contains zero to near-zero coefficients for undesirable wavelengths. The method is based on Tikhonov regularization with the model 1-norm (TR1) and is applied to simulated and near-infrared (NIR) spectral data. Models are also formed from wavelength subsets determined by the standard method of stepwise regression (SWR). Harmonious (bias/variance tradeoff) and parsimonious considerations are compared with and without wavelength selection for principal component regression (PCR), ridge regression (RR), partial least squares (PLS), and multiple linear regression (MLR). Results show that TR1 models generally contain large baseline regions of near-zero coefficients, thereby essentially achieving built-in wavelength selection. For example, wavelengths with spectral interferences and/or poor signal-to-noise ratios obtain near zero regression coefficients. Results often improve with TR1 models, compared to full wavelength PCR, RR, and PLS models. The SWR subset results are similar to those for the TR1 models using the NIR data and worse with the simulated spectral situations. In general, wavelength selection improves prediction accuracy at a sacrifice to a potential increase in variance and the parsimony remains nearly equivalent compared to full wavelength models. New insights gained from the reported studies provide useful guidelines on when to use full wavelengths or use wavelength selection methods. Specifically, when a small number of large wavelength effects (good sensitivity and selectivity) exist, subset selection by SWR (with caution) and TR1 do well. With a small to moderate number of large to moderate sized wavelength effects, TR1 is better. Lastly, when a large number of small effects are present, full wavelengths with the methods of PCR, RR, or PLS are best.     

Neuroadaptations in ionotropic and metabotropic glutamate receptor mRNA produced by cocaine treatment

The expression of glutamate receptor/subunit mRNAs was examined 3 weeks after discontinuing 1 week of daily injections of saline or cocaine. The level of mRNA for GluR1-4, NMDAR1, and mGluR5 receptors was measured with in situ hybridization and RT-PCR. In nucleus accumbens, acute cocaine treatment significantly reduced the mRNA level for GluR3, GluR4, and NMDAR1 subunits, whereas repeated cocaine reduced the level for GluR3 mRNA. Acute cocaine treatment also reduced the NMDAR1 mRNA level in dorsolateral striatum and ventral tegmental area. In prefrontal cortex, repeated cocaine treatment significantly increased the level of GluR2 mRNA. The GluR2 mRNA level was not changed by acute or repeated cocaine in any other brain regions examined. Repeated cocaine treatment also significantly increased mGluR5 mRNA levels in nucleus accumbens shell and dorsolateral striatum. Functional properties of the ionotropic glutamate receptors are determined by subunit composition. In addition, metabotropic glutamate receptors can modulate synaptic transmission and the response to stimulation of ionotropic receptors. Thus, the observed changes in levels of AMPA and NMDA receptor subunits and the mGluR5 metabotropic receptor may alter excitatory neurotransmission in the mesocorticolimbic dopamine system, which could play a significant role in the enduring biochemical and behavioral effects of cocaine.     

Synthesis of A356 Al–high-Ca fly ash composites by pressure infiltration technique and their characterization

Eight types of A356 Al–fly ash composites were produced by pressure infiltration of high-Ca lignite fly ash. This type of ash was used for the first time in Al-composites synthesis, and particularly by liquid metal infiltration techniques. After examining mineralogy and chemistry, specific, narrow ash size fractions were used for the synthesis of composites, and properties linked to microstructure and wear strength of the materials. The effect of using ground ash particles on the microstructure and tribological performance of the composites was also investigated. It was concluded that using fine, high-Ca ash particles can improve the properties of composites, and that using ash particles in a ground form can better facilitate the production process of MMCs.     

Substance P in the ventral pallidum: projection from the ventral striatum, and electrophysiological and behavioral consequences of pallidal substance P

1. Piroximone was administered orally (p.o.) and intravenously (i.v.) to male Beagle dog. In vitro, piroximone was incubated with dog liver microsomes. 2. Piroximone was metabolized in vivo to five metabolites (1-5) representing approximately 20% of the total administered dose. 3. The parent drug and its metabolites were totally eliminated in urine. 4. Reduced piroximone (piroximole), representing approximately 10% of the administered dose, was identified as the major metabolic product in vivo. 5. In vitro, piroximone was metabolized by dog liver microsomes to isonicotinic acid (1) and piroximole (4), with the same ratio as in vivo (1:4 = 0.2). The Michaelis-Menten parameters were determined for piroximole formation and were: Kmapp = 733 microM and Vmax app = 232 pmol/mg protein/min. 6. Comparison of the pharmacokinetics of piroximone and piroximole revealed that both compounds were very well absorbed (F = 93 +/- 7 and 89 +/- 8% respectively), slightly distributed (Vd app = 0.78 +/- 0.04 and 1.02 +/- 0.09 l/kg p.o., and 0.95 +/- 0.05 and 0.76 +/- 0.13 1/kg i.v. respectively) and excreted into urine to the same extent (UEx = 54.7 +/- 1.2 and 53.2 +/- 12.6% p.o., and 59.1 +/- 5.3 and 51.2 +/- 5.7% i.v. respectively), except that the clearance of piroximone was two-fold higher than that observed for piroximole (ClT = 7.77 +/- 1.35 and 4.12 +/- 0.44 ml/min/kg p.o., and 7.68 +/- 1.25 and 4.06 +/- 0.51 ml/min/kg i.v. respectively).     

Characterization of the frequency stability of frequency-hopping sources

A method that applies time-domain stability measures, hitherto used only in the context of fixed-frequency sources, to the characterization of frequency-hopped sources is introduced. This method has been successfully applied to the testing of a novel practical frequency-hopping synthesizer. It is shown that despite the fact that hopping sources exhibit an orders-of-magnitude larger amount of short term noise than stable free-running frequency sources, some of the techniques used to estimate the noise of stable sources are also applicable to hopping voltage-controlled oscillator systems.