Multiple Imputation of missing values in exploratory factor analysis of multidimensional scales: estimating latent trait scores

ABSTRACT Researchers frequently have to analyze scales in which some participants have failed to respond to some items. In this paper we focus on the exploratory factor analysis of multidimensional scales (i.e., scales that consist of a number of subscales) where each subscale is made up of a number of Likert-type items, and the aim of the analysis is to estimate participants' scores on the corresponding latent traits. We propose a new approach to deal with missing responses in such a situation that is based on (1) multiple imputation of non-responses and (2) simultaneous rotation of the imputed datasets. We applied the approach in a real dataset where missing responses were artificially introduced following a real pattern of non-responses, and a simulation study based on artificial datasets. The results show that our approach (specifically, Hot-Deck multiple imputation followed of Consensus Promin rotation) was able to successfully compute factor score estimates even for participants that have missing data.

Missing value imputation in longitudinal measures of alcohol consumption

Attrition in longitudinal studies can lead to biased results. The study is motivated by the unexpected observation that alcohol consumption decreased despite increased availability, which may be due to sample attrition of heavy drinkers. Several imputation methods have been proposed, but rarely compared in longitudinal studies of alcohol consumption. The imputation of consumption level measurements is computationally particularly challenging due to alcohol consumption being a semi-continuous variable (dichotomous drinking status and continuous volume among drinkers), and the non-normality of data in the continuous part. Data come from a longitudinal study in Denmark with four waves (2003-2006) and 1771 individuals at baseline. Five techniques for missing data are compared: Last value carried forward (LVCF) was used as a single, and Hotdeck, Heckman modelling, multivariate imputation by chained equations (MICE), and a Bayesian approach as multiple imputation methods. Predictive mean matching was used to account for non-normality, where instead of imputing regression estimates, "real" observed values from similar cases are imputed. Methods were also compared by means of a simulated dataset. The simulation showed that the Bayesian approach yielded the most unbiased estimates for imputation. The finding of no increase in consumption levels despite a higher availability remained unaltered. Copyright (C) 2011 John Wiley & Sons, Ltd.

Modelagem estocástica do índice de radiação ultravioleta na costa leste do nordeste do Brasil

The increase in ultraviolet radiation (UV) at surface, the high incidence of non-melanoma skin cancer (NMSC) in coast of Northeast of Brazil (NEB) and reduction of total ozone were the motivation for the present study. The overall objective was to identify and understand the variability of UV or Index Ultraviolet Radiation (UV Index) in the capitals of the east coast of the NEB and adjust stochastic models to time series of UV index aiming make predictions (interpolations) and forecasts / projections (extrapolations) followed by trend analysis. The methodology consisted of applying multivariate analysis (principal component analysis and cluster analysis), Predictive Mean Matching method for filling gaps in the data, autoregressive distributed lag (ADL) and Mann-Kendal. The modeling via the ADL consisted of parameter estimation, diagnostics, residuals analysis and evaluation of the quality of the predictions and forecasts via mean squared error and Pearson correlation coefficient. The research results indicated that the annual variability of UV in the capital of Rio Grande do Norte (Natal) has a feature in the months of September and October that consisting of a stabilization / reduction of UV index because of the greater annual concentration total ozone. The increased amount of aerosol during this period contributes in lesser intensity for this event. The increased amount of aerosol during this period contributes in lesser intensity for this event. The application of cluster analysis on the east coast of the NEB showed that this event also occurs in the capitals of Paraiba (João Pessoa) and Pernambuco (Recife). Extreme events of UV in NEB were analyzed from the city of Natal and were associated with absence of cloud cover and levels below the annual average of total ozone and did not occurring in the entire region because of the uneven spatial distribution of these variables. The ADL (4, 1) model, adjusted with data of the UV index and total ozone to period 2001-2012 made a the projection / extrapolation for the next 30 years (2013-2043) indicating in end of that period an increase to the UV index of one unit (approximately), case total ozone maintain the downward trend observed in study period

La representación como estrategia de resolución de problemas.

Elaborar materiales para el entrenamiento de los alumnos en estrategias de resolución de problemas. Establecer pautas que permitan utilizar estas estrategias como instrumento metodológico en el área de Matemáticas. Comprobar las posibles diferencias entre sexos. Demostrar la transferibilidad de estas estrategias a otras situaciones. Comprobar las posibles variaciones de CI en la población experimental. Hipótesis de trabajo: utilizar la estrategia de la representación mejora los resultados en la resolución de problemas. 206 sujetos alumnos de octavo de EGB y tercero de ESO perteneceientes a cinco centros del municipio de Cartagena: 4 públicos y 1 privado concertado localizados en zonas urbanas y semiurbanas; resultado de la aplicación de un procedimiento (Mean Matching) para igualar las distribuciones de las variables perturbadoras en los grupos experimental y de control. Varones 38 por ciento y mujeres 62 por ciento. Edad media: 13.4 años. Variables independientes: presencia o ausencia de tratamiento (dicotómica), sexo (dicotómica), grupo al que pertenece: experimental/control (dicotómica). Variable dependiente: resultados obtenidos por los alumnos en las pruebas de Matemáticas, Ciencias Sociales y Ciencias Naturales, teniendo en cuenta el sexo y la utilización o no de la estrategia (variable de frecuencia). Pruebas de Matemáticas, Ciencias Sociales y Ciencias Naturales utilizadas en fase pretest y posttest. Protocolo TEA-2 en las dos fases. Escala de actitud hacia las Matemáticas. Paquete estadísico SYSTAT TABLES. Análisis descriptivo de las variables sexo y grupo: histograma de frecuencias. Análisis de correlaciones utilizando el estadístico Mantel-Haenszel. Análisis de varianza. Análisis descriptivo de la escala de actitud. 1. El uso de estrategias de representación aumenta la capacidad de los alumnos para resolver problemas, sobre todo en el área de Matemáticas y tanto para hombres como para mujeres; 2. Existe transferencia de estrategias para la resolución de problemas a otras áreas y por tanto, una generalización; 3. El entrenamiento ha incidido de forma distinta en hombres y mujeres; 4. Aumento del CI general; 5. El cambio de actitud hacia las Matemáticas no ha podido ser demostrado. El estudio contiene un programa de instrucción que puede enmarcarse dentro del paradigma instruccional del aprendizaje ya que no pretende crear cambios de naturaleza cuantitativa o cualitativa en los sujetos, sino facilitar las condiciones para que se adquieran las estrategias típicas que un experto aplica cuando soluciona un problema.

Comparison of selective genotyping strategies for prediction of breeding values in a population undergoing selection

Genomewide marker information can improve the reliability of breeding value predictions for young selection candidates in genomic selection. However, the cost of genotyping limits its use to elite animals, and how such selective genotyping affects predictive ability of genomic selection models is an open question. We performed a simulation study to evaluate the quality of breeding value predictions for selection candidates based on different selective genotyping strategies in a population undergoing selection. The genome consisted of 10 chromosomes of 100 cM each. After 5,000 generations of random mating with a population size of 100 (50 males and 50 females), generation G(0) (reference population) was produced via a full factorial mating between the 50 males and 50 females from generation 5,000. Different levels of selection intensities (animals with the largest yield deviation value) in G(0) or random sampling (no selection) were used to produce offspring of G(0) generation (G(1)). Five genotyping strategies were used to choose 500 animals in G(0) to be genotyped: 1) Random: randomly selected animals, 2) Top: animals with largest yield deviation values, 3) Bottom: animals with lowest yield deviations values, 4) Extreme: animals with the 250 largest and the 250 lowest yield deviations values, and 5) Less Related: less genetically related animals. The number of individuals in G(0) and G(1) was fixed at 2,500 each, and different levels of heritability were considered (0.10, 0.25, and 0.50). Additionally, all 5 selective genotyping strategies (Random, Top, Bottom, Extreme, and Less Related) were applied to an indicator trait in generation G(0), and the results were evaluated for the target trait in generation G(1), with the genetic correlation between the 2 traits set to 0.50. The 5 genotyping strategies applied to individuals in G(0) (reference population) were compared in terms of their ability to predict the genetic values of the animals in G(1) (selection candidates). Lower correlations between genomic-based estimates of breeding values (GEBV) and true breeding values (TBV) were obtained when using the Bottom strategy. For Random, Extreme, and Less Related strategies, the correlation between GEBV and TBV became slightly larger as selection intensity decreased and was largest when no selection occurred. These 3 strategies were better than the Top approach. In addition, the Extreme, Random, and Less Related strategies had smaller predictive mean squared errors (PMSE) followed by the Top and Bottom methods. Overall, the Extreme genotyping strategy led to the best predictive ability of breeding values, indicating that animals with extreme yield deviations values in a reference population are the most informative when training genomic selection models.

Rainfall spatial predictions: a two-part model and its assessment

Spatial prediction of hourly rainfall via radar calibration is addressed. The change of support problem (COSP), arising when the spatial supports of different data sources do not coincide, is faced in a non-Gaussian setting; in fact, hourly rainfall in Emilia-Romagna region, in Italy, is characterized by abundance of zero values and right-skeweness of the distribution of positive amounts. Rain gauge direct measurements on sparsely distributed locations and hourly cumulated radar grids are provided by the ARPA-SIMC Emilia-Romagna. We propose a three-stage Bayesian hierarchical model for radar calibration, exploiting rain gauges as reference measure. Rain probability and amounts are modeled via linear relationships with radar in the log scale; spatial correlated Gaussian effects capture the residual information. We employ a probit link for rainfall probability and Gamma distribution for rainfall positive amounts; the two steps are joined via a two-part semicontinuous model. Three model specifications differently addressing COSP are presented; in particular, a stochastic weighting of all radar pixels, driven by a latent Gaussian process defined on the grid, is employed. Estimation is performed via MCMC procedures implemented in C, linked to R software. Communication and evaluation of probabilistic, point and interval predictions is investigated. A non-randomized PIT histogram is proposed for correctly assessing calibration and coverage of two-part semicontinuous models. Predictions obtained with the different model specifications are evaluated via graphical tools (Reliability Plot, Sharpness Histogram, PIT Histogram, Brier Score Plot and Quantile Decomposition Plot), proper scoring rules (Brier Score, Continuous Rank Probability Score) and consistent scoring functions (Root Mean Square Error and Mean Absolute Error addressing the predictive mean and median, respectively). Calibration is reached and the inclusion of neighbouring information slightly improves predictions. All specifications outperform a benchmark model with incorrelated effects, confirming the relevance of spatial correlation for modeling rainfall probability and accumulation.

Forecasting with serially correlated regression models

In this article we investigate the asymptotic and finite-sample properties of predictors of regression models with autocorrelated errors. We prove new theorems associated with the predictive efficiency of generalized least squares (GLS) and incorrectly structured GLS predictors. We also establish the form associated with their predictive mean squared errors as well as the magnitude of these errors relative to each other and to those generated from the ordinary least squares (OLS) predictor. A large simulation study is used to evaluate the finite-sample performance of forecasts generated from models using different corrections for the serial correlation.

Closed-loop control of mean arterial blood pressure during surgery with alfentanil: clinical evaluation of a novel model-based predictive controller

BACKGROUND: In contrast to hypnosis, there is no surrogate parameter for analgesia in anesthetized patients. Opioids are titrated to suppress blood pressure response to noxious stimulation. The authors evaluated a novel model predictive controller for closed-loop administration of alfentanil using mean arterial blood pressure and predicted plasma alfentanil concentration (Cp Alf) as input parameters. METHODS: The authors studied 13 healthy patients scheduled to undergo minor lumbar and cervical spine surgery. After induction with propofol, alfentanil, and mivacurium and tracheal intubation, isoflurane was titrated to maintain the Bispectral Index at 55 (+/- 5), and the alfentanil administration was switched from manual to closed-loop control. The controller adjusted the alfentanil infusion rate to maintain the mean arterial blood pressure near the set-point (70 mmHg) while minimizing the Cp Alf toward the set-point plasma alfentanil concentration (Cp Alfref) (100 ng/ml). RESULTS: Two patients were excluded because of loss of arterial pressure signal and protocol violation. The alfentanil infusion was closed-loop controlled for a mean (SD) of 98.9 (1.5)% of presurgery time and 95.5 (4.3)% of surgery time. The mean (SD) end-tidal isoflurane concentrations were 0.78 (0.1) and 0.86 (0.1) vol%, the Cp Alf values were 122 (35) and 181 (58) ng/ml, and the Bispectral Index values were 51 (9) and 52 (4) before surgery and during surgery, respectively. The mean (SD) absolute deviations of mean arterial blood pressure were 7.6 (2.6) and 10.0 (4.2) mmHg (P = 0.262), and the median performance error, median absolute performance error, and wobble were 4.2 (6.2) and 8.8 (9.4)% (P = 0.002), 7.9 (3.8) and 11.8 (6.3)% (P = 0.129), and 14.5 (8.4) and 5.7 (1.2)% (P = 0.002) before surgery and during surgery, respectively. A post hoc simulation showed that the Cp Alfref decreased the predicted Cp Alf compared with mean arterial blood pressure alone. CONCLUSION: The authors' controller has a similar set-point precision as previous hypnotic controllers and provides adequate alfentanil dosing during surgery. It may help to standardize opioid dosing in research and may be a further step toward a multiple input-multiple output controller.

Predictive Validity of Waterlow Scale for Pressure Ulcer Development Risk in Hospitalized Patients

AIM: We sought to evaluate the predictive validity of the Waterlow Scale in hospitalized patients. SUBJECTS AND SETTING: The study was conducted at a general private hospital with 220 beds and a mean time of hospitalization of 7.4 days and a mean occupation rate of approximately 80%. Adult patients with a Braden Scale score of 18 or less and a Waterlow Scale score of 16 or more were studied. The sample consisted of 98 patients with a mean age of 71.1 +/- 15.5 years. METHODS: Skin assessment and scoring by using the Waterlow and Braden scales were completed on alternate days. Patients were examined at least 3 times to be considered for analysis. The data were submitted to sensitivity and specificity analysis by using receiver operating characteristic (ROC) curves and positive (+LR) and negative (-LR) likelihood ratios. RESULTS: The cutoff scores were 17, 20, and 20 in the first, second, and third assessment, respectively. Sensitivity was 71.4%, 85.7%, and 85.7% and specificity was 67.0%, 40.7%, and 32.9%, respectively. Analysis of the area under the ROC curve revealed good accuracy (0.64, 95% confidence interval [CI]: 0.35-0.93) only for the cutoff score 17 in the first assessment. The results also showed probabilities of 14%, 10%, and 9% for the development of pressure ulcer when the test results were positive (+LR) and of 3% (-LR) when the test results were negative for the cutoff scores in the first, second, and third assessment, respectively. CONCLUSION: The Waterlow Scale achieved good predictive validity in predicting pressure ulcer in hospitalized patients when a cutoff score of 17 was used in the first assessment.

The critical power function is dependent on the duration of the predictive exercise tests chosen

The linear relationship between work accomplished (W-lim) and time to exhaustion (t(lim)) can be described by the equation: W-lim = a + CP.t(lim). Critical power (CP) is the slope of this line and is thought to represent a maximum rate of ATP synthesis without exhaustion, presumably an inherent characteristic of the aerobic energy system. The present investigation determined whether the choice of predictive tests would elicit significant differences in the estimated CP. Ten female physical education students completed, in random order and on consecutive days, five art-out predictive tests at preselected constant-power outputs. Predictive tests were performed on an electrically-braked cycle ergometer and power loadings were individually chosen so as to induce fatigue within approximately 1-10 mins. CP was derived by fitting the linear W-lim-t(lim) regression and calculated three ways: 1) using the first, third and fifth W-lim-t(lim) coordinates (I-135), 2) using coordinates from the three highest power outputs (I-123; mean t(lim) = 68-193 s) and 3) using coordinates from the lowest power outputs (I-345; mean t(lim) = 193-485 s). Repeated measures ANOVA revealed that CPI123 (201.0 +/- 37.9W) > CPI135 (176.1 +/- 27.6W) > CPI345 (164.0 +/- 22.8W) (P < 0.05). When the three sets of data were used to fit the hyperbolic Power-t(lim) regression, statistically significant differences between each CP were also found (P < 0.05). The shorter the predictive trials, the greater the slope of the W-lim-t(lim) regression; possibly because of the greater influence of 'aerobic inertia' on these trials. This may explain why CP has failed to represent a maximal, sustainable work rate. The present findings suggest that if CP is to represent the highest power output that an individual can maintain for a very long time without fatigue then CP should be calculated over a range of predictive tests in which the influence of aerobic inertia is minimised.

Matching occupation and self: Does matching theory adequately model children's thinking?

The present exploratory-descriptive cross-national study focused on the career development of 11- to 14-yr.-old children, in particular whether they can match their personal characteristics with their occupational aspirations. Further, the study explored whether their matching may be explained in terms of a fit between person and environment using Holland's theory as an example. Participants included 511 South African and 372 Australian children. Findings relate to two items of the Revised Career Awareness Survey that require children to relate personal-social knowledge to their favorite occupation. Data were analyzed in three stages using descriptive statistics, i.e., mean scores, frequencies, and percentage agreement. The study indicated that children perceived their personal characteristics to be related to their occupational aspirations. However, how this matching takes place is not adequately accounted for in terms of a career theory such as that of Holland.

PREDICTIVE FACTORS FOR GAIT IN FEMORAL TRANSTROCHANTERIC FRACTURES

Objective: The objective of this study was to identify variables that could predict the quality of gait in patients with transtrochanteric femoral fractures after treatment. Materials and Methods: Hospitalized patients diagnosed with transtrochanteric femoral fractures were selected between September/2005 and August/2006 and followed-up for 6 months after the trauma date. An observational prospective study was conducted to assess the quality of gait 3 and 6 months after fracture in 31 patients (13 males and 18 females). The mean age was 76 +/- 2,7. Results: Seven patients (22,6%) passed away during the follow-up period. The patients with associated fractures or with four or more co-morbidities showed a worse quality of gait after 6 months. Patients without orthopaedic complications or who got partial weight load prior to 30 days showed a better performance. Conclusion: The quantification of predictive gait indexes allows us to propose new treatment approaches consistently to the different realities showed by each group of patients.

Predictive factors for the development of visual loss in patients with pituitary macroadenomas and for visual recovery after optic pathway decompression

Objective: To investigate clinical and MRI findings that are predictive of both visual loss in patients with pituitary adenomas and visual recovery after treatment. Design: Cohort study. Participants: Thirty patients (60 eyes) with pituitary adenoma. Methods: Patients underwent neuro-ophthalmic examination and MRI before and after optic chiasm decompression. Visual field (VF) was assessed using the mean deviation in standard automated perimetry (SAP) and temporal mean defect, the average of 22 temporal values of the total deviation plot. Tumour size was measured on sagittal and coronal cuts. Results: Visual loss was found in 47 eyes; 35 had optic atrophy (subtle in 9, moderate in 14, and severe in 12). Before treatment, the average SAP mean deviation and temporal mean defect were -11.78 (SD 8.56) dB and -18.66 (SD 11.20) dB, respectively. The chiasm was 17.3 (SD 6.2, range 10-34) mm above the reference line on the sagittal and 21.8 (SD 8.3, range 12-39) mm on the coronal images. Tumour size correlated with the severity of VF defect. VF improvement occurred in 80% of eyes after treatment. The degree of optic atrophy, visual loss, and tumour size were significantly associated with improvement after treatment. Conclusions: The best predictive factor for visual loss was tumour size, and factors related to visual recovery were the degree of optic atrophy, the severity of VF defect, and the tumour size. Diagnosing pituitary adenomas before optic atrophy becomes severe may be related to a better prognosis in such patients.

MRI based diffusion and perfusion predictive model to estimate stroke evolution

In this study we present a novel automated strategy for predicting infarct evolution, based on MR diffusion and perfusion images acquired in the acute stage of stroke. The validity of this methodology was tested on novel patient data including data acquired from an independent stroke clinic. Regions-of-interest (ROIs) defining the initial diffusion lesion and tissue with abnormal hemodynamic function as defined by the mean transit time (MTT) abnormality were automatically extracted from DWI/PI maps. Quantitative measures of cerebral blood flow (CBF) and volume (CBV) along with ratio measures defined relative to the contralateral hemisphere (r(a)CBF and r(a)CBV) were calculated for the MTT ROIs. A parametric normal classifier algorithm incorporating these measures was used to predict infarct growth. The mean r(a)CBF and r(a)CBV values for eventually infarcted MTT tissue were 0.70 +/-0.19 and 1.20 +/-0.36. For recovered tissue the mean values were 0.99 +/-0.25 and 1.87 +/-0.71, respectively. There was a significant difference between these two regions for both measures (P

Prognostic modelling of breast cancer patients: a benchmark of predictive models with external validation

Dissertação apresentada para obtenção do Grau de Doutor em Engenharia Electrotécnica e de Computadores – Sistemas Digitais e Percepcionais pela Universidade Nova de Lisboa, Faculdade de Ciências e Tecnologia