Robust regression in biological assay: application to the evaluation of alternative experimental techniques.

Robust Huber type regression and testing of linear hypotheses are adapted to statistical analysis of parallel line and slope ratio assays. They are applied in the evaluation of results of several experiments carried out in order to compare and validate alternatives to animal experimentation based on embryo and cell cultures. Computational procedures necessary for the application of robust methods of analysis used the conversational statistical package ROBSYS. Special commands for the analysis of parallel line and slope ratio assays have been added to ROBSYS.

Robust accelerated failure time regression

Robust estimators for accelerated failure time models with asymmetric (or symmetric) error distribution and censored observations are proposed. It is assumed that the error model belongs to a log-location-scale family of distributions and that the mean response is the parameter of interest. Since scale is a main component of mean, scale is not treated as a nuisance parameter. A three steps procedure is proposed. In the first step, an initial high breakdown point S estimate is computed. In the second step, observations that are unlikely under the estimated model are rejected or down weighted. Finally, a weighted maximum likelihood estimate is computed. To define the estimates, functions of censored residuals are replaced by their estimated conditional expectation given that the response is larger than the observed censored value. The rejection rule in the second step is based on an adaptive cut-off that, asymptotically, does not reject any observation when the data are generat ed according to the model. Therefore, the final estimate attains full efficiency at the model, with respect to the maximum likelihood estimate, while maintaining the breakdown point of the initial estimator. Asymptotic results are provided. The new procedure is evaluated with the help of Monte Carlo simulations. Two examples with real data are discussed.

Does history of falling predict length of stay in post-scute geriatric rehabilitation?

Objectives: To investigate the associations between falls before¦hospital admission, falls during hospitalization, and length of stay in¦elderly people admitted to post-acute geriatric rehabilitation.¦Method: History of falling in the previous 12 months before admission¦was recorded among 249 older persons (mean age 82.3 ± 7.4 years,¦69.1% women) consecutively admitted to post-acute rehabilitation. Data¦on medical, functional and cognitive status were collected upon¦admission. Falls during hospitalization and length of stay were recorded¦at discharge.¦Results: Overall, 92 (40.4%) patients reported no fall in the 12 months¦before admission; 63(27.6%) reported 1 fall, and 73 (32.0%) reported¦multiple falls. Previous falls occurrence (one or more falls) was¦significantly associated with in-stay falls (19.9% of previous fallers fell¦during the stay vs 7.6% in patients without history of falling, P = .01),¦and with a longer length of stay (22.4 ± 10.1 days vs 27.1 ± 14.3 days,¦P = .01). In multivariate robust regression controlling for gender, age,¦functional and cognitive status, history of falling remained significantly¦associated with longer rehabilitation stay (2.8 days more than non¦fallers in single fallers, p = .05, and 3.3 days in multiple fallers, p = .0.1).¦Conclusion: History of falling in the 12 months prior to post acute¦geriatric rehabilitation is independently associated with a longer¦rehabilitation length of stay. Previous fallers also have an increased risk¦of falling during rehabilitation stay. This suggests that hospital fall¦prevention measures should particularly target these high risk patients.

Does history of falling predict length of stay in post-acute geriatric rehabilitation?

Objectives:To investigate the associations between falls before hospital¦admission, falls during hospitalization, and length of stay in elderly¦people admitted to post-acute geriatric rehabilitation. Method: History¦of falling in the previous 12 months before admission was recorded¦among 249 older persons (mean age 82.3±7.4 years, 69.1% women)¦consecutively admitted to post-acute rehabilitation. Data on medical,¦functional and cognitive status were collected upon admission. Falls¦during hospitalization and length of stay were recorded at discharge.¦Results: Overall, 92 (40.4%) patients reported no fall in the 12 months¦before admission; 63(27.6%) reported 1 fall, and 73(32.0%) reported¦multiple falls. Previous falls occurrence (one or more falls) was significantly¦associated with in-stay falls (19.9% of previous fallers fell¦during the stay vs 7.6% in patients without history of falling, P=.01),¦and with a longer length of stay (22.4 ± 10.1 days vs 27.1 ± 14.3 days,¦P=.01). In multivariate robust regression controlling for gender, age,¦functional and cognitive status, history of falling remained significantly¦associated with longer rehabilitation stay (2.8 days more in single fallers,¦p=.05, and 3.3 days more in multiple fallers, p=.0.1, compared to¦non-fallers). Conclusion: History of falling in the 12 months prior to¦post acute geriatric rehabilitation is independently associated with a¦longer rehabilitation length of stay. Previous fallers have also an¦increased risk of falling during rehabilitation stay. This suggests that¦hospital fall prevention measures should particularly target these high¦riskpatients.

Red cell distribution width does not predict stroke severity or functional outcome.

INTRODUCTION: Red cell distribution width was recently identified as a predictor of cardiovascular and all-cause mortality in patients with previous stroke. Red cell distribution width is also higher in patients with stroke compared with those without. However, there are no data on the association of red cell distribution width, assessed during the acute phase of ischemic stroke, with stroke severity and functional outcome. In the present study, we sought to investigate this relationship and ascertain the main determinants of red cell distribution width in this population. METHODS: We used data from the Acute Stroke Registry and Analysis of Lausanne for patients between January 2003 and December 2008. Red cell distribution width was generated at admission by the Sysmex XE-2100 automated cell counter from ethylene diamine tetraacetic acid blood samples stored at room temperature until measurement. An χ(2) -test was performed to compare frequencies of categorical variables between different red cell distribution width quartiles, and one-way analysis of variance for continuous variables. The effect of red cell distribution width on severity and functional outcome was investigated in univariate and multivariate robust regression analysis. Level of significance was set at 95%. RESULTS: There were 1504 patients (72±15·76 years, 43·9% females) included in the analysis. Red cell distribution width was significantly associated to NIHSS (β-value=0·24, P=0·01) and functional outcome (odds ratio=10·73 for poor outcome, P<0·001) at univariate analysis but not multivariate. Prehospital Rankin score (β=0·19, P<0·001), serum creatinine (β=0·008, P<0·001), hemoglobin (β=-0·009, P<0·001), mean platelet volume (β=0·09, P<0·05), age (β=0·02, P<0·001), low ejection fraction (β=0·66, P<0·001) and antihypertensive treatment (β=0·32, P<0·001) were independent determinants of red cell distribution width. CONCLUSIONS: Red cell distribution width, assessed during the early phase of acute ischemic stroke, does not predict severity or functional outcome.

Differential effect of Charlson index and its components on in-hospital mortality and health costs

Introduction: The Charlson index (Charlson, 1987) is a commonly used comorbidity index in outcome studies. Still, the use of different weights makes its calculation cumbersome, while the sum of its components (comorbidities) is easier to compute. In this study, we assessed the effects of 1) the Charlson index adapted for the Swiss population and 2) the sum of its components (number of comorbidities, maximum 15) on a) in-hospital deaths and b) cost of hospitalization. Methods: Anonymous data was obtained from the administrative database of the department of internal medicine of the Lausanne University Hospital (CHUV). All hospitalizations of adult (>=18 years) patients occurring between 2003 and 2011 were included. For each hospitalization, the Charlson index and the number of comorbidities were calculated. Analyses were conducted using Stata. Results: Data from 32,741 hospitalizations occurring between 2003 and 2011 was analyzed. On bivariate analysis, both the Charlson index and the number of comorbidities were significantly and positively associated with in hospital death. Conversely, multivariate adjustment for age, gender and calendar year using Cox regression showed that the association was no longer significant for the number of comorbidities (table). On bivariate analysis, hospitalization costs increased both with Charlson index and with number of comorbidities, but the increase was much steeper for the number of comorbidities (figure). Robust regression after adjusting for age, gender, calendar year and duration of hospital stay showed that the increase in one comorbidity led to an average increase in hospital costs of 321 CHF (95% CI: 272 to 370), while the increase in one score point of the Charlson index led to a decrease in hospital costs of 49 CHF (95% CI: 31 to 67). Conclusion: Charlson index is better than the number of comorbidities in predicting in-hospital death. Conversely, the number of comorbidities significantly increases hospital costs.

Hypertension is an independent predictor of mean platelet volume in patients with acute ischaemic stroke.

Introduction: Mean platelet volume (MPV) was shown to be significantly increased in patients with acute ischaemic stroke, especially in non-lacunar strokes. Moreover, some studies concluded that increased MPV is related to poor functional outcome after ischaemic stroke, although this association is still controversial. However, the determinants of MPV in patients with acute ischaemic stroke have never been investigated. Subjects and methods: We recorded the main demographic, clinical and laboratory data of consecutive patients with acute (admitted within 24 h after stroke onset) ischaemic stroke admitted in our Neurology Service between January 2003 and December 2008. MPV was generated at admission by the Sysmex XE-2100 automated cell counter (Sysmex Corporation, Kobe, Japan) from ethylenediaminetetraacetic acid blood samples stored at room temperature until measurement. The association of these parameters with MPV was investigated in univariate and multivariate analysis. Results: A total of 636 patients was included in our study. The median MPV was 10.4 ± 0.82 fL. In univariate analysis, glucose (β= 0.03, P= 0.05), serum creatinine (β= 0.002, P= 0.02), haemoglobin (β= 0.009, P < 0.001), platelet count (β=-0.002, P < 0.001) and history of arterial hypertension (β= 0.21, P= 0.005) were found to be significantly associated with MPV. In multivariate robust regression analysis, only hypertension and platelet count remained as independent determinants of MPV. Conclusions: In patients with acute ischaemic stroke, platelet count and history of hypertension are the only determinants of MPV.

Mesure des plis cutanés et masse adipeuse corporelle: une précision de l'excès de poids dans l'enquête épidémiologique

By definition, obesity corresponds to the presence of a mass of fatty tissue that is excessive with respect to the body mass. Body fat can be calculated in terms of age and sex by measuring the skinfold thickness in several different places. During the MONICA project, the survey of cardiovascular risk factor prevalence enabled us to measure the thickness of four skinfolds (biceps, triceps, subscapular, suprailiac) in 263 inhabitants of Lausanne (125 men, 138 women). In men aged 25-34, 21 +/- 5% of the body mass was composed of fat, in women 29 +/- 4%. The proportion of fat increases to 31 +/- 7% in men and 41 +/- 6% in women aged 55-64. A robust regression allows body fat to be simply expressed in terms of the body mass index. This allows us to confirm the validity of this index for evaluating the degree of obesity during an epidemiological study.

J-shaped association between serum glucose and functional outcome in acute ischemic stroke.

BACKGROUND AND PURPOSE: Hyperglycemia after stroke is associated with larger infarct volume and poorer functional outcome. In an animal stroke model, the association between serum glucose and infarct volume is described by a U-shaped curve with a nadir ≈7 mmol/L. However, a similar curve in human studies was never reported. The objective of the present study is to investigate the association between serum glucose levels and functional outcome in patients with acute ischemic stroke. METHODS: We analyzed 1446 consecutive patients with acute ischemic stroke. Serum glucose was measured on admission at the emergency department together with multiple other metabolic, clinical, and radiological parameters. National Institutes of Health Stroke Scale (NIHSS) score was recorded at 24 hours, and Rankin score was recorded at 3 and 12 months. The association between serum glucose and favorable outcome (Rankin score ≤2) was explored in univariate and multivariate analysis. The model was further analyzed in a robust regression model based on fractional polynomial (-2-2) functions. RESULTS: Serum glucose is independently correlated with functional outcome at 12 months (OR, 1.15; P=0.01). Other predictors of outcome include admission NIHSS score (OR, 1.18; P<0001), age (OR, 1.06; P<0.001), prestroke Rankin score (OR, 20.8; P=0.004), and leukoaraiosis (OR, 2.21; P=0.016). Using these factors in multiple logistic regression analysis, the area under the receiver-operator characteristic curve is 0.869. The association between serum glucose and Rankin score at 12 months is described by a J-shaped curve with a nadir of 5 mmol/L. Glucose values between 3.7 and 7.3 mmol/L are associated with favorable outcome. A similar curve was generated for the association of glucose and 24-hour NIHSS score, for which glucose values between 4.0 and 7.2 mmol/L are associated with a NIHSS score <7. Discussion-Both hypoglycemia and hyperglycemia are dangerous in acute ischemic stroke as shown by a J-shaped association between serum glucose and 24-hour and 12-month outcome. Initial serum glucose values between 3.7 and 7.3 mmol/L are associated with favorable outcome.

Robust Nonlinear Mapping of Soil Contamination Using Support Regression

Spatial data analysis mapping and visualization is of great importance in various fields: environment, pollution, natural hazards and risks, epidemiology, spatial econometrics, etc. A basic task of spatial mapping is to make predictions based on some empirical data (measurements). A number of state-of-the-art methods can be used for the task: deterministic interpolations, methods of geostatistics: the family of kriging estimators (Deutsch and Journel, 1997), machine learning algorithms such as artificial neural networks (ANN) of different architectures, hybrid ANN-geostatistics models (Kanevski and Maignan, 2004; Kanevski et al., 1996), etc. All the methods mentioned above can be used for solving the problem of spatial data mapping. Environmental empirical data are always contaminated/corrupted by noise, and often with noise of unknown nature. That's one of the reasons why deterministic models can be inconsistent, since they treat the measurements as values of some unknown function that should be interpolated. Kriging estimators treat the measurements as the realization of some spatial randomn process. To obtain the estimation with kriging one has to model the spatial structure of the data: spatial correlation function or (semi-)variogram. This task can be complicated if there is not sufficient number of measurements and variogram is sensitive to outliers and extremes. ANN is a powerful tool, but it also suffers from the number of reasons. of a special type ? multiplayer perceptrons ? are often used as a detrending tool in hybrid (ANN+geostatistics) models (Kanevski and Maignank, 2004). Therefore, development and adaptation of the method that would be nonlinear and robust to noise in measurements, would deal with the small empirical datasets and which has solid mathematical background is of great importance. The present paper deals with such model, based on Statistical Learning Theory (SLT) - Support Vector Regression. SLT is a general mathematical framework devoted to the problem of estimation of the dependencies from empirical data (Hastie et al, 2004; Vapnik, 1998). SLT models for classification - Support Vector Machines - have shown good results on different machine learning tasks. The results of SVM classification of spatial data are also promising (Kanevski et al, 2002). The properties of SVM for regression - Support Vector Regression (SVR) are less studied. First results of the application of SVR for spatial mapping of physical quantities were obtained by the authorsin for mapping of medium porosity (Kanevski et al, 1999), and for mapping of radioactively contaminated territories (Kanevski and Canu, 2000). The present paper is devoted to further understanding of the properties of SVR model for spatial data analysis and mapping. Detailed description of the SVR theory can be found in (Cristianini and Shawe-Taylor, 2000; Smola, 1996) and basic equations for the nonlinear modeling are given in section 2. Section 3 discusses the application of SVR for spatial data mapping on the real case study - soil pollution by Cs137 radionuclide. Section 4 discusses the properties of the modelapplied to noised data or data with outliers.

Robust response transformations based on optimal prediction

Nonlinear regression problems can often be reduced to linearity by transforming the response variable (e.g., using the Box-Cox family of transformations). The classic estimates of the parameter defining the transformation as well as of the regression coefficients are based on the maximum likelihood criterion, assuming homoscedastic normal errors for the transformed response. These estimates are nonrobust in the presence of outliers and can be inconsistent when the errors are nonnormal or heteroscedastic. This article proposes new robust estimates that are consistent and asymptotically normal for any unimodal and homoscedastic error distribution. For this purpose, a robust version of conditional expectation is introduced for which the prediction mean squared error is replaced with an M scale. This concept is then used to develop a nonparametric criterion to estimate the transformation parameter as well as the regression coefficients. A finite sample estimate of this criterion based on a robust version of smearing is also proposed. Monte Carlo experiments show that the new estimates compare favorably with respect to the available competitors.

Robust volume-targeted balanced steady-state free-precession coronary magnetic resonance angiography in a breathhold at 3.0 Tesla: a reproducibility study.

BACKGROUND: Transient balanced steady-state free-precession (bSSFP) has shown substantial promise for noninvasive assessment of coronary arteries but its utilization at 3.0 T and above has been hampered by susceptibility to field inhomogeneities that degrade image quality. The purpose of this work was to refine, implement, and test a robust, practical single-breathhold bSSFP coronary MRA sequence at 3.0 T and to test the reproducibility of the technique. METHODS: A 3D, volume-targeted, high-resolution bSSFP sequence was implemented. Localized image-based shimming was performed to minimize inhomogeneities of both the static magnetic field and the radio frequency excitation field. Fifteen healthy volunteers and three patients with coronary artery disease underwent examination with the bSSFP sequence (scan time = 20.5 ± 2.0 seconds), and acquisitions were repeated in nine subjects. The images were quantitatively analyzed using a semi-automated software tool, and the repeatability and reproducibility of measurements were determined using regression analysis and intra-class correlation coefficient (ICC), in a blinded manner. RESULTS: The 3D bSSFP sequence provided uniform, high-quality depiction of coronary arteries (n = 20). The average visible vessel length of 100.5 ± 6.3 mm and sharpness of 55 ± 2% compared favorably with earlier reported navigator-gated bSSFP and gradient echo sequences at 3.0 T. Length measurements demonstrated a highly statistically significant degree of inter-observer (r = 0.994, ICC = 0.993), intra-observer (r = 0.894, ICC = 0.896), and inter-scan concordance (r = 0.980, ICC = 0.974). Furthermore, ICC values demonstrated excellent intra-observer, inter-observer, and inter-scan agreement for vessel diameter measurements (ICC = 0.987, 0.976, and 0.961, respectively), and vessel sharpness values (ICC = 0.989, 0.938, and 0.904, respectively). CONCLUSIONS: The 3D bSSFP acquisition, using a state-of-the-art MR scanner equipped with recently available technologies such as multi-transmit, 32-channel cardiac coil, and localized B0 and B1+ shimming, allows accelerated and reproducible multi-segment assessment of the major coronary arteries at 3.0 T in a single breathhold. This rapid sequence may be especially useful for functional imaging of the coronaries where the acquisition time is limited by the stress duration and in cases where low navigator-gating efficiency prohibits acquisition of a free breathing scan in a reasonable time period.

Geomodelling of a fluvial system with semi-supervised support vector regression

Fluvial deposits are a challenge for modelling flow in sub-surface reservoirs. Connectivity and continuity of permeable bodies have a major impact on fluid flow in porous media. Contemporary object-based and multipoint statistics methods face a problem of robust representation of connected structures. An alternative approach to model petrophysical properties is based on machine learning algorithm ? Support Vector Regression (SVR). Semi-supervised SVR is able to establish spatial connectivity taking into account the prior knowledge on natural similarities. SVR as a learning algorithm is robust to noise and captures dependencies from all available data. Semi-supervised SVR applied to a synthetic fluvial reservoir demonstrated robust results, which are well matched to the flow performance