Lower Total Error Research Articles

Bayesian unit-root tests for Stochastic Volatility models

In this article, we consider Bayesian inference procedures to test for a unit root in Stochastic Volatility (SV) models. Unit-root tests for the persistence parameter of the SV models, based on the Bayes Factor (BF), have been recently introduced in the literature. In contrast, we propose a flexible class of priors that is non-informative over the entire support of the persistence parameter (including the non-stationarity region). In addition, we show that our model fitting procedure is computationally efficient (using the software WinBUGS). Finally, we show that our proposed test procedures have good frequentist properties in terms of achieving high statistical power, while maintaining low total error rates. We illustrate the above features of our method by extensive simulation studies, followed by an application to a real data set on exchange rates.

Total body water estimations in healthy men and women using bioimpedance spectroscopy: a deuterium oxide comparison

BackgroundTotal body water (TBW) estimations have been used to estimate body composition, particularly fat-free mass, to aid in nutritional interventions, and to monitor hydration status. In the past, bioimpedance spectroscopy (BIS) devices have been used to estimate TBW. Previous investigations have examined the validity of the XiTRON 4000B (XiTRON Technologies) BIS device for estimating TBW. Recently, a new BIS device (Imp™ SFB7) has become available, claiming greater precision when estimating TBW. The Imp™ SFB7 (SFB7) is based on similar BIS principles, while offering increased portability and a greater range of frequencies when compared to older devices, such as the XiTRON 4000B (4000B). The purpose of this study was to examine the validity of the SFB7 for estimating total body water in healthy college-age men and women compared to the 4000B and deuterium oxide (D2O).MethodsTwenty-eight Caucasian men and women (14 men, 14 women; 24 ± 4 yrs; 174.6 ± 8.7 cm; 72.80 ± 17.58 kg) had their TBW estimated by the SFB7, the 4000B, and D2O.ResultsBoth BIS devices produced similar standard error of estimate (SEE) and r values (SFB7, SEE = 2.12L, r = 0.98; 4000B, SEE = 2.99L, r = 0.96) when compared to D2O, though a significant constant error (CE) was detected for the 4000B (2.26L, p ≤ 0.025). The 4000B produced a larger total error (TE) and CE (TE = 3.81L, CE = 2.26L) when compared to the SFB7 (TE = 2.21L, CE = -0.09L). Additionally, the limits of agreement were larger for the 4000B (-3.88 to 8.39L) than the SFB7 (-4.50 to 4.31L). These results were consistent when sex was analyzed separately, though women produced lower SEE and TE values for both devices.ConclusionThe 4000B and SFB7 are valid BIS devices when compared to D2O to estimate TBW in college-age Caucasian men and women. Furthermore, the new SFB7 device displayed greater precision in comparison to the 4000B, which may decrease the error when estimating TBW on an individual basis.

Validity of the BOD POD for Assessing Body Composition in Athletic High School Boys

The purpose of this study was to validate the percentage of body fat (%BF) values estimated from the BOD POD (BP) with those obtained from hydrostatic weighing (HW) in athletic American high school boys. Additionally, the %BF values measured via near-infrared interactance (NIR), bioelectrical impedance (BIA), and skinfold (SF) were compared to HW to determine the validity of these measures. Thirty white boys (mean age +/- SD = 15.8 +/- 1.0 years) who where currently participating in organized sports volunteered to have their %BF estimated. Measurements were obtained from NIR, BP, BIA, and SF in random order and concluded with HW. The findings from the present study indicated that the NIR and BIA instruments produced significant (P < 0.008) constant error (CE) and total error (TE) values that were too large to be of practical value (TE > 4.0%BF). The BP produced a significantly (P < 0.008) higher CE with acceptable TE values compared to HW, but compared to all three SF estimations, the BP TE values were higher. Two of the SF equations were nonsignificant (P > 0.008) and had the lowest TE values compared to HW. These data suggest that the BP can produce acceptable body fat measures for athletic white boys but is not superior to estimates made by the SF equations used in this study.

Percent body fat estimations in college women using field and laboratory methods: a three-compartment model approach.

BackgroundMethods used to estimate percent body fat can be classified as a laboratory or field technique. However, the validity of these methods compared to multiple-compartment models has not been fully established. This investigation sought to determine the validity of field and laboratory methods for estimating percent fat (%fat) in healthy college-age women compared to the Siri three-compartment model (3C).MethodsThirty Caucasian women (21.1 ± 1.5 yrs; 164.8 ± 4.7 cm; 61.2 ± 6.8 kg) had their %fat estimated by BIA using the BodyGram™ computer program (BIA-AK) and population-specific equation (BIA-Lohman), NIR (Futrex® 6100/XL), a quadratic (SF3JPW) and linear (SF3WB) skinfold equation, air-displacement plethysmography (BP), and hydrostatic weighing (HW).ResultsAll methods produced acceptable total error (TE) values compared to the 3C model. Both laboratory methods produced similar TE values (HW, TE = 2.4%fat; BP, TE = 2.3%fat) when compared to the 3C model, though a significant constant error (CE) was detected for HW (1.5%fat, p ≤ 0.006). The field methods produced acceptable TE values ranging from 1.8 – 3.8 %fat. BIA-AK (TE = 1.8%fat) yielded the lowest TE among the field methods, while BIA-Lohman (TE = 2.1%fat) and NIR (TE = 2.7%fat) produced lower TE values than both skinfold equations (TE > 2.7%fat) compared to the 3C model. Additionally, the SF3JPW %fat estimation equation resulted in a significant CE (2.6%fat, p ≤ 0.007).ConclusionData suggest that the BP and HW are valid laboratory methods when compared to the 3C model to estimate %fat in college-age Caucasian women. When the use of a laboratory method is not feasible, NIR, BIA-AK, BIA-Lohman, SF3JPW, and SF3WB are acceptable field methods to estimate %fat in this population.

Analysis and Design of Iterative Learning Control Strategies for UPS Inverters

In this paper, two iterative-learning-based control methods for uninterruptible power system (UPS) inverters, viz., direct iterative learning controller (direct ILC) and hybrid iterative learning controller (hybrid ILC), are proposed. In both methods, a zero-phase filter that is designed in the frequency domain is applied in order to provide compensation for the resonant peak in the system model to ensure error convergence. Furthermore, a "forgetting factor" is introduced in both control algorithms to increase the robustness of the scheme against measurement noise, initialization error, and/or variation of system dynamics due to any parameter drift. In the direct ILC method, ILC is combined with the feedforward of the reference to improve dynamic performance. This method is shown to be capable of achieving very high steady-state performance although its dynamic response was not very good. In the hybrid ILC method, ILC is combined with a proportional-derivative controller aside from the reference feedforward to improve the dynamic response further. Experimental results show that both proposed controllers can achieve very low total harmonic distortion and fast error convergence under different loads while using only one sensor. The proposed direct ILC is an effective solution for UPS products where high-quality steady-state output voltage is more important than fast dynamic response, while hybrid ILC can result in excellent steady-state performance with a more improved dynamic response than direct ILC

CROSS-VALIDATION OF VENTILATORY THRESHOLD PREDICTION EQUATIONS ON AEROBICALLY TRAINED MEN AND WOMEN

The purpose of this investigation was to determine the validity of the non-exercise-based equations of Davis et al. (13), Jones et al. (20), and Neder et al. (30) for estimating the ventilatory threshold (VT) in samples of aerobically trained men and women. One hundred and forty-four aerobically trained men (mean +/- SD age, 41.0 +/- 11.6 years; N = 83) and women (37.1 +/- 9.0 years, N = 61) performed a maximal incremental test to determine VO2max and observed VT on a cycle ergometer. The observed VT was determined by gas exchange measurements using the V-slope method (VCO2/VO2) in conjunction with analyses of the ventilatory equivalents (i.e., minute ventilation VE/VO2 and VE/VCO2) and end-tidal gas tensions (i.e., P(ET)O2 and P(ET)CO2) for oxygen and carbon dioxide. The predicted VT values from 14 equations were compared to the observed VT values by examining the constant error (CE), standard error of estimate (SEE), Pearson correlation coefficient (r), and total error (TE). The results of this investigation indicated that all 14 equations resulted in significant (p < 0.008) CE values ranging from 1.13 to 1.72 L x min(-1) for the men and from 0.58 to 1.12 L x min(-1) for the women. Furthermore, the SEE, r, and TE values ranged from 0.37 to 0.54, from 0.36 to 0.53, and from 0.68 to 1.81 L x min(-1), respectively. The lowest TE values for the men and women represented 45 and 36% of the mean of the observed VT values, respectively. The results of this study indicated that the errors associated with all 14 equations were too large to be of practical value for estimating VT in aerobically trained men and women.

Classifying class I and class II compounds by hydrophobicity and hydrogen bonding descriptors.

The successful development of quantitative structure-activity relationships (QSARs) and the prediction of toxicity based on QSARs depend on the correct classification of the mechanism of toxic action of chemical compounds. The toxicity mechanism of a compound can be determined by a few existing methods, such as studying the chemical structure of the compound for certain substructures and assigning a mechanism accordingly. However, these methods are less reliable for compounds with complex structures and are usually complicated. A novel descriptor-based method employing a multivariate statistical technique--discriminant analysis--was developed in this study for the classification of class I and class II compounds. The discriminating variables used in discriminant analysis were the hydrophobicity descriptor log(K(ow)) and the hydrogen bonding descriptors E(LUMO), E(HOMO), Q(+), and Q(-). Because only numerical values of these descriptors are required for this method, which can be calculated, no additional experimental work was required for the toxicity mechanism classification of new compounds. A nonlinear discriminant function was generated that could not be expressed explicitly. Assessing the predicting ability of the discriminant function by the cross-validation method showed that a low total error rate, 4.2% was achieved.

VALIDITY OF BIOELECTRICAL IMPEDANCE EQUATIONS FOR ESTIMATING FAT-FREE WEIGHT IN LEAN FEMALES

The present study examined the validity of bioelectrical impedance (BIA) equations for estimating fat-free weight (FFW) in lean males (X +/- SD = 9.1 +/- 2.2% fat) by comparing the estimates with values obtained from underwater weighing. Sixty-eight Caucasian male volunteers served as subjects. Cross-validation analyses included examination of the constant error (CE), standard error of the estimate (SEE), r, and total error (TE). The results indicated that the equations of Oppliger et al. (16), which resulted in small TE (1.70 kg) and CE (-0.02 kg) values, most accurately estimated FFW. Simple linear regression showed that FFW was more highly correlated with body weight (BW) (r = 0.98, P < 0.0001) and resulted in a lower SEE (1.68 kg) than either height2/resistance (Ht2/R) (r = 0.81, P < 0.0001; SEE = 5.12 kg) or the independent variable (weight x resistance)/height2 [WR/Ht2] utilized by the manufacturer of the BIA analyzer (r = 0.15, P > 0.05; SEE = 8.59 kg). Multiple regression showed that when WR/Ht2, Ht2/R, resistance, body mass index, Ht2, and/or Ht was added to the prediction equation, which utilized BW alone, they accounted for less than 1% additional variance and reduced the SEE by < or = 0.16 kg. The results indicated that BW alone estimated FFW as accurately as any of the BIA equations in lean males.

Predicting forage indigestible NDF from lignin concentration.

We used chemical composition and in vitro digestibility data from temperate and tropical forages to develop relationships between indices of lignification and forage indigestible NDF. Neutral detergent fiber indigestibility increased nonlinearly as the lignin concentration of the NDF increased. Differences in estimated indigestible NDF using equations developed for a specific forage class (C3 and C4 grasses and legumes) were small and are probably not biologically significant when compared to those estimated from a common equation. Selected equations were compared with the Cornell Net Carbohydrate and Protein System (CNCPS) for the prediction of ADG. The linear equation (2.4 times NDF lignin content) used by the CNCPS and the Beef NRC had some of the largest errors due to mean bias. A log-log model [4.37 x (lignin/NDF)(.84)] provided the best combination of low total prediction error, low mean bias, and minimal error due to regression bias when permanganate lignin was used. A similar equation based on sulfuric acid lignin [6.17 x (lignin/NDF)(.77)] also met the above criteria. These equations then were evaluated with the CNCPS model against animal growth data from diets ranging in forage quality. Regardless of the equation used for predicting unavailable fiber, the CNCPS underpredicted daily gain, with mean biases ranging from -.10 to -.22 kg/d. Regression bias ranged from .13 to .14 kg/d and the coefficients differed from unity (P = .0001). The new equations gave numerically lower energy allowable ADG by steers compared to the linear equation currently used by the CNCPS model. The estimates were lower due to a higher predicted indigestible NDF, which resulted in a lower estimated forage energy value.

Validity of near-infrared interactance for estimating relative body fat in female high school gymnasts.

The present investigation examined the validity of near-infrared interactance (NIR) estimates of relative body fat (% fat) from the Futrex-5000 (F5000), Futrex-5000A (F5000 A), and Futrex-1000 (F1000) instruments. Ninety-eight female high school gymnasts (X age+/-SD = 15.7+/-1.2 yr) participated in this investigation. Subsamples were used to cross-validate the F5000 (n = 52), F5000A (n = 46), and F1000 (n = 89) instruments. The NIR % fat estimates were validated against criterion % fat from underwater weighing (UWW) using the adult conversion constants of Brozek et al. (6) (UWWB) and the female age-specific constants of Lohman (23) (UWWL). The cross-validation statistical analysis included examination of the constant error (CE), standard error of estimate (SEE), r, and total error (TE). In addition, full-model multiple regression analyses were used to predict UWWB or UWWL from body weight (BW) and height (HT). BW and HT were correlated with % fat at R = 0.65-0.70, while the validity coefficients for the NIR instruments ranged from r = 0.40-0.78. The F5000 resulted in nonsignificant CE values (-0.3 % fat vs UWWB and 1.5 % fat vs UWWL; p > 0.008) as well as the lowest TE values (TE = 3.1 % fat vs UWWB and 4.0 % fat vs UWWL). All other NIR estimates of % fat resulted in TE values > or = 6.3 % fat. In addition, for all NIR instruments there were negative correlations for the plots of the CE versus the mean of predicted and criterion (UWWB and UWWL) % fat. Therefore, the present findings indicated that the F5000 provided more accurate estimates of % fat than the F5000A or F1000 instruments, but may underestimate the desired minimal body weight for female gymnasts at the low end of the % fat distribution.

USE OF A SINGLE SITE SKINFOLD EQUATION TO ESTIMATE% FAT IN PERSONS WITH AIDS1057

Determination of%fat may be an important clinical parameter in patients with AIDS-associated wasting. Therefore, we studied the accuracy of using skinfold thickness for estimating%fat in this population. METHODS Subjects were 51 clinically stable men with AIDS-associated wasting (>10% weight loss). Subjects' heights and weights were recorded, and skinfold thicknesses were obtained from the biceps, triceps, subscapula, and suprailium. Percent fat was estimated using the four site skinfold equation of Dumin and Womersley (1974). Dual energy x-ray absorptiometry (DXA) was used as the criterion%fat measure. We compared the four site equation to%fat values obtained from DXA in a validation sample (n=36). Also, we derived an equation using subjects' height, weight, and skinfold thicknesses. The four site equation and our derived equation were then used to estimate%fat in a cross validation sample (n=25). RESULTS There were no significant differences in age (35.0±6.3 yr), height (176.2±6.7 cm), weight(64.3±7.9 kg), or sum of four skinfolds (33.2±9.9 mm) between the validation and cross validation groups. Percent fat as measured by DXA in the validation sample was 14.8±5.3%, compared to 15.8±4.1% with the four site equation. The correlation was r=0.60, with a standard error of estimate (SEE) of 4.2%, and a total error (TE) of 4.4% for the skinfold estimate. Our derived equation; [% Fat = 1.113 * (triceps skinfold) + 0.834 * (BMI) - 10.792], yielded 14.8±4.0%fat, with an r= 0.77, SEE of 3.4%, and TE of 3.3%. Using our derived equation on the cross validation sample resulted in%fat values almost identical to DXA(13.2±4.4 vs 13.8±3.0%), while the four site equation overestimated at 17.2±3.7%. Further, our derived equation yielded a higher correlation (r=0.65 vs 0.49), lower SEE (3.3 vs 3.8%fat), and lower TE(3.4 vs 5.7%fat) than the four site skinfold equation. CONCLUSION We found a derived equation to estimate%fat in persons with AIDS to be superior to a traditional, four site equation. The need for only height, weight, and triceps skinfold thickness emphasizes the new equation's utility in a clinic setting.

Evaluation of Alternative Equations for Prediction of Intake for Holstein Dairy Cows

Six prediction equations for dry matter intake (DMI) were evaluated for accuracy with independent data. The equations were selected based on ease of parameter measurement and practical on-farm use. The equations were assessed for accuracy of predicting individual weekly DMI for primiparous (n = 105) and multiparous (n = 136) cows; three-fourths of these cows were supplemented with a sustained-release form of bovine somatotropin (bST). Large variations in accuracy were identified across the six prediction equations for effects of parity and bST. Prediction accuracy of all equations for cows in wk 1 to 24 of lactation was better for primiparous cows than for multiparous cows. Precision of prediction equations was poor for cows in wk 8 through 12 of lactation and for cows in >40 wk of lactation. The equation for DMI with the best accuracy measured by a low total lactation mean square prediction error was the modified equation of the National Research Council: DMI (kilograms per day) = –0.293 + 0.372 × fat-corrected milk (kilograms per day) + 0.0968 × body weight0.75 (kilograms). However, the overall mean bias (predicted minus observed) of the prediction of weekly DMI of a single cow was high for all equations, including the modified equation of the National Research Council. For wk 2, 4, 8, 10, and 20 of lactation, the mean bias for the modified equation was +6, +3.4, –1.3, –2.1, and –2.8 kg/d. The accuracy of prediction was lower for cows treated biweekly with bST. High yielding cows and cows treated biweekly with bST had higher milk yields in relation to body weight, and standardized prediction equations for DMI were less accurate.

Within-subject transformations of PET regional cerebral blood flow data: ANCOVA, ratio, and Z-score adjustments on empirical data

In the statistical analysis of PET rCBF data, it has become routine to use a mathematical transformation to reduce individual differences in global metabolic rate and increase the power of statistical tests. Two methods of adjustment have been proposed: the ratio adjustment, dividing rCBF for a brain region by the whole-brain average CBF, and the ANCOVA adjustment based on a regression model. We compared these two transformations on the empirically derived PET data sets to assess which of these transforms worked best. Within-subject Z-scores were also considered as an additional transformation technique. Comparisons between the three transform techniques, and the untransformed raw data, were made on the detected significant differences, the size of the statistical effects, and the relative reduction in error variance. Two data sets were considered: 1) a visual perception study comparing object and spatial vision; 2) a data set from a parametric visual working memory study. There was no striking difference in the detected significant differences between any of the three transformed data sets and all transformed data sets detected more differences than the raw data set. There was little difference between the transformed data sets in terms of effect sizes, with the Z-score data set showing slightly lower total error variance that either the ratio-adjusted or ANCOVA-adjusted data sets. It was concluded that PET data sets should be closely examined for task-related changes in global CBF and subject heterogeneities (outliers) since these factors will influence the outcome of image-based statistical comparison more than the particular data transformation.

The validity of visual estimations of percent body fat in lean males

The present study compared the validity of visual estimations of percent fat (% fat) in lean males (mean +/- SD = 9.6 +/- 2.3 % fat) to the validity of bioelectrical impedance analysis (BIA) and skinfold equations. Thirty-five Caucasian male volunteers (mean +/- SD = 23 +/- 5 yr; range = 19-40) served as subjects. Visual estimations of % fat were performed by two experienced male raters. The validity (compared to underwater weighing) for each procedure was determined by examining the constant error (CE), standard error of the estimate (SEE), r, and total error (TE). The results indicated that rater 1 (TE = 2.3% fat) could visually estimate % fat as accurately as the skinfold equations (TE = 2.4% fat). However, based on low TE, SEE, and CE values as well as considerable variability (mean difference = 2.7% fat) between the % fat estimates of the two raters, skinfold equations are recommended over visual inspection and BIA (TE = 5.0% fat) for estimating % fat in lean males.

Fluorometric method for determination of vitamin B6 in soya bean

The simple fluorometric method for determination of added pyridoxine-HCl in enriched food, based on oxidation of pyridoxine to 4-pyridoxic acid by means of KMnO4, has been modified for determination of total B6 in soya bean. In comparison to the other fluorometric procedures, the proposed method is the simplest one, rapid and easy to perform, demanding the least time for oxidation and not using hazardous chemicals (KMnO4 instead of KCN). Good recovery, low relative standard deviation and low total error allow the proposed procedure to be included into the group of excellent methods.

A Family of Highly Accurate Interpolation Functions for Magnitude-Mode Fourier Transform Spectroscopy

A new family of highly efficient interpolating functions, the KCe functions, KCe( ω) = ( aω2 + bω + c) e, where e is the exponent, is developed for three-point frequency interpolation of discrete, magnitude-mode, apodized Fourier transform spectra. The family is characterized by high interpolation accuracy and ease of implementation. Various members of the family can be generated by varying the exponent. Prior work from this laboratory indicated that the parabola is the interpolating function of choice for interpolation of discrete, apodized magnitude spectra. We show here that, compared to parabolic interpolation, KCe interpolation typically gives residual systematic errors which are lower by between one and two orders of magnitude. These systematic errors are analytically derived and the efficacy of interpolation is rigorously examined as a function of the KCe exponent, the number of zero-fillings, the amount of damping in the transient, and the window function used to apodize the spectrum. For Hanning-apodized spectra, the KC5.5 function gives the lowest residual systematic errors, which are typically 15 times less than those remaining after parabolic interpolation. Similarly, the KC6.6 function is optimal for Hamming-apodized spectra (22 times better than parabolic interpolation) and the KC9.5 function is optimal for Blackman-Harris-apodized spectra (80 times better than parabolic interpolation). By extrapolation from other optimal KCe functions, we estimate that the optimal KCe function for interpolation of Kaiser-Bessel-apodized spectra is KC12.5. Analytical formulae for propagation of random errors in spectral intensity into random errors in interpolated frequency are derived for parabolic interpolation and for KCe interpolation. These error propagation formulae give random errors which are inversely proportional to the SNR of the spectrum. These formulae are evaluated with the appropriate KCe exponent for each of the Hanning, the Hamming, and the Blackman-Harris windows. In all cases we find that the random error is essentially independent of both window type and interpolation scheme. While zero-filling prior to interpolation reduces the residual systematic frequency interpolation error, it increases the random frequency error. The increase in random error with higher levels of zero-filling is explained. Because the random errors are proportional to noise level, the optimal number of zero-fillings varies with SNR. If the apodizing window is chosen to match the dynamic range of the spectrum, as we have previously recommended, then the systematic error for KCe interpolation of non-zero-filled spectra is so low that the overall error is dominated by the random error. In this case, KCe interpolation is, for all intents and purposes, exact. Since the random error is minimized by no zero-filling, the lowest overall error will be achieved by a combination of no zero-filling and KCe interpolation. In constrast, the minimum total error for parabolic interpolation is achieved by interpolation of the once-zero-filled spectrum. A further advantage of KCe interpolation, over and above its lower total error, is that KCe interpolation obviates the need for zero-filling.

A Model to Evaluate Alternative Methods of Defibrillation Threshold Determination

The voltage (or, equivalently, energy) at which defibrillation occurs for a specific episode of fibrillation can be represented by specifying or estimating the probability of successful defibrillation for each voltage or energy. This relation of voltage to probability is the probability function. For a series of attempts, the probability function predicts the frequency with which defibrillation will occur at a given voltage. By defining the defibrillation threshold (DFT) as the voltage at which the probability function takes on a specific value, say 50%, a method of defibrillation threshold determination can be evaluated by how accurately and precisely it estimates the "true" defibrillation threshold. By utilizing estimated probability functions from animals and humans, the relative performance of different methods of defibrillation threshold determination can be evaluated. Three methods were evaluated using published animal data and human clinical data: (1) stepping down to the first voltage that fails; (2) stepping up to the first voltage that succeeds; and (3) doing both 1 and 2 and averaging. In all cases, Method 3 has lower total error. Definitions of defibrillation threshold other than the 50% level can also be evaluated in this fashion.

Estimation of minimal wrestling weight using measures of body build and body composition.

This study evaluated field methods for establishing minimal body weight in wrestlers based upon estimates of fat-free body weight (FFB): 239 high school wrestlers were measured for height, weight, 13 diameters, 24 circumferences, and 7 skinfolds, as well as for underwater weight to determine body density and resultant FFB. To estimate FFB, anthropometric models were used which were based upon diameters, circumferences, and/or skinfolds. Correlations between actual and predicted FFB were high, ranging from r = 0.89 to 0.92 for the models employing diameters and r = 0.96 for those employing circumferences, which respectively increased to r = 0.96 and r = 0.97 when body weight was included in theses models. The skinfold-based and mixed model estimates of FFB were r = 0.97 and r = 0.91 to 0.97, respectively. The corresponding total error (TE) in these estimates for the diameter-based models ranged from 6.54 to 3.80 kg (inclusion of body weight further reduced TE to 2.77 kg). Among the circumference models, TE ranged from 2.50 to 2.33 kg, with inclusion of body weight accounting for the lower value. Skinfold-based estimates of FFB demonstrated a TE level ranging from 2.48 to 2.33 kg. The lowest TE (2.32 kg) was demonstrated by a mixed model which included skinfolds, diameter, height, and body weight. Cross-validation of these new equations revealed the lowest TE values (less than 2.2 kg) for the skinfold and the mixed model equations, thereby identifying highly valid field techniques for determining the present level of FFB, which can serve as the basis upon which minimal body weight estimates can be made.

Forecasting interstate migration with limited data: a demographic-economic approach.

"The limitations of available migration data preclude a time-series approach of modeling interstate migration [in the United States]. The method presented here combines aspects of the demographic and economic approaches to forecasting migration in a manner compatible with existing data. Migration rates are modeled to change in response to changes in economic conditions. When applied to resently constructed data on migration based on income tax returns and then compared to standard demographic projections, the demographic-economic approach has a 20% lower total error in forecasting net migration by state for cohorts of labor-force age."

A Comparison of Medication Errors Under the University of Kentucky Unit Dose System and Traditional Drug Distribution Systems in Four Hospitals

The University of Kentucky Hospital has operated a hospital-wide unit dose system of dispensing medications for more than 4 years. 2 community and 2 university hospitals still using the traditional medication systems were used in a comparative study to ascertain which system leads to more medication errors. The criteria for hospital selection are discussed. Medication error audits were performed on selected medical nursing units in each of the 5 hospitals in the study. The auditing methodology is described. Medication errors which were measured included dosage errors timing errors and omissions of medication. The University of Kentucky Hospital was found to have a statistically significantly lower total error rate combining errors of omission and errors of commission. The 4 hospitals chosen for comparative purposes differed among themselves in the number and type of errors found. Information handling was found to be a major source of drug error. At the University of Kentucky Hospital the pharmacists were found to play an active role in the drug distribution process as opposed to the other hospitals where the pharmacists were relegated to a superficial role.