Multivariate Analysis for Animal Selection in Experimental Research

Background: Several researchers seek methods for the selection of homogeneous groups of animals in experimental studies, a fact justified because homogeneity is an indispensable prerequisite for casualization of treatments. The lack of robust methods that comply with statistical and biological principles is the reason why researchers use empirical or subjective methods, influencing their results. Objective: To develop a multivariate statistical model for the selection of a homogeneous group of animals for experimental research and to elaborate a computational package to use it. Methods: The set of echocardiographic data of 115 male Wistar rats with supravalvular aortic stenosis (AoS) was used as an example of model development. Initially, the data were standardized, and became dimensionless. Then, the variance matrix of the set was submitted to principal components analysis (PCA), aiming at reducing the parametric space and at retaining the relevant variability. That technique established a new Cartesian system into which the animals were allocated, and finally the confidence region (ellipsoid) was built for the profile of the animals’ homogeneous responses. The animals located inside the ellipsoid were considered as belonging to the homogeneous batch; those outside the ellipsoid were considered spurious. Results: The PCA established eight descriptive axes that represented the accumulated variance of the data set in 88.71%. The allocation of the animals in the new system and the construction of the confidence region revealed six spurious animals as compared to the homogeneous batch of 109 animals. Conclusion: The biometric criterion presented proved to be effective, because it considers the animal as a whole, analyzing jointly all parameters measured, in addition to having a small discard rate.

Hepatitis C prevalence and risk factors in hemodialysis patients in Central Brazil: a survey by polymerase chain reaction and serological methods

An hemodialysis population in Central Brazil was screened by polymerase chain reaction (PCR) and serological methods to assess the prevalence of hepatitis C virus (HCV) infection and to investigate associated risk factors. All hemodialysis patients (n=428) were interviewed in eight dialysis units in Goiânia city. Blood samples were collected and serum samples screened for anti-HCV antibodies by an enzyme-linked immunosorbent assay (ELISA). Positive samples were retested for confirmation with a line immunoassay (LIA). All samples were also tested for HCV RNA by the PCR. An overall prevalence of 46.7% (CI 95%: 42-51.5) was found, ranging from 20.7% (CI 95%: 8.8-38.1) to 90.4% (CI 95%: 79.9-96.4) depending on the dialysis unit. Of the 428 patients, 185 were found to be seropositive by ELISA, and 167 were confirmed positive by LIA, resulting in an anti-HCV prevalence of 39%. A total of 131 patients were HCV RNA-positive. HCV viremia was present in 63.5% of the anti-HCV-positive patients and in 10.3% of the anti-HCV-negative patients. Univariate analysis of risk factors showed that the number of previous blood transfusions, transfusion of blood before mandatory screening for anti-HCV, length of time on hemodialysis, and treatment in multiple units were associated with HCV positivity. However, multivariate analysis revealed that blood transfusion before screening for anti-HCV and length of time on hemodialysis were significantly associated with HCV infection in this population. These data suggest that nosocomial transmission may play a role in the spread of HCV in the dialysis units studied. In addition to anti-HCV screening, HCV RNA detection is necessary for the diagnosis of HCV infection in hemodialysis patients.

Multivariate QST-FST comparisons: a neutrality test for the evolution of the g matrix in structured populations.

Neutrality tests in quantitative genetics provide a statistical framework for the detection of selection on polygenic traits in wild populations. However, the existing method based on comparisons of divergence at neutral markers and quantitative traits (Q(st)-F(st)) suffers from several limitations that hinder a clear interpretation of the results with typical empirical designs. In this article, we propose a multivariate extension of this neutrality test based on empirical estimates of the among-populations (D) and within-populations (G) covariance matrices by MANOVA. A simple pattern is expected under neutrality: D = 2F(st)/(1 - F(st))G, so that neutrality implies both proportionality of the two matrices and a specific value of the proportionality coefficient. This pattern is tested using Flury's framework for matrix comparison [common principal-component (CPC) analysis], a well-known tool in G matrix evolution studies. We show the importance of using a Bartlett adjustment of the test for the small sample sizes typically found in empirical studies. We propose a dual test: (i) that the proportionality coefficient is not different from its neutral expectation [2F(st)/(1 - F(st))] and (ii) that the MANOVA estimates of mean square matrices between and among populations are proportional. These two tests combined provide a more stringent test for neutrality than the classic Q(st)-F(st) comparison and avoid several statistical problems. Extensive simulations of realistic empirical designs suggest that these tests correctly detect the expected pattern under neutrality and have enough power to efficiently detect mild to strong selection (homogeneous, heterogeneous, or mixed) when it is occurring on a set of traits. This method also provides a rigorous and quantitative framework for disentangling the effects of different selection regimes and of drift on the evolution of the G matrix. We discuss practical requirements for the proper application of our test in empirical studies and potential extensions.

Fusion of data sets in multivariate linear regression with errors-in-variables

We consider the application of normal theory methods to the estimation and testing of a general type of multivariate regressionmodels with errors--in--variables, in the case where various data setsare merged into a single analysis and the observable variables deviatepossibly from normality. The various samples to be merged can differ on the set of observable variables available. We show that there is a convenient way to parameterize the model so that, despite the possiblenon--normality of the data, normal--theory methods yield correct inferencesfor the parameters of interest and for the goodness--of--fit test. Thetheory described encompasses both the functional and structural modelcases, and can be implemented using standard software for structuralequations models, such as LISREL, EQS, LISCOMP, among others. An illustration with Monte Carlo data is presented.

Asymptotic robustness in multi-sample analysis of multivariate linear relations

Standard methods for the analysis of linear latent variable models oftenrely on the assumption that the vector of observed variables is normallydistributed. This normality assumption (NA) plays a crucial role inassessingoptimality of estimates, in computing standard errors, and in designinganasymptotic chi-square goodness-of-fit test. The asymptotic validity of NAinferences when the data deviates from normality has been calledasymptoticrobustness. In the present paper we extend previous work on asymptoticrobustnessto a general context of multi-sample analysis of linear latent variablemodels,with a latent component of the model allowed to be fixed across(hypothetical)sample replications, and with the asymptotic covariance matrix of thesamplemoments not necessarily finite. We will show that, under certainconditions,the matrix $\Gamma$ of asymptotic variances of the analyzed samplemomentscan be substituted by a matrix $\Omega$ that is a function only of thecross-product moments of the observed variables. The main advantage of thisis thatinferences based on $\Omega$ are readily available in standard softwareforcovariance structure analysis, and do not require to compute samplefourth-order moments. An illustration with simulated data in the context ofregressionwith errors in variables will be presented.

Use of NIR spectroscopy and multivariate process spectra calibration methodology for pharmaceutical solid samples analysis

Accomplish high quality of final products in pharmaceutical industry is a challenge that requires the control and supervision of all the manufacturing steps. This request created the necessity of developing fast and accurate analytical methods. Near infrared spectroscopy together with chemometrics, fulfill this growing demand. The high speed providing relevant information and the versatility of its application to different types of samples lead these combined techniques as one of the most appropriated. This study is focused on the development of a calibration model able to determine amounts of API from industrial granulates using NIR, chemometrics and process spectra methodology.

Eliminació d'artefactes en EGG mitjançant l'ús de la Multivariate Empirical Mode Decomposition

La tècnica de l’electroencefalograma (EEG) és una de les tècniques més utilitzades per estudiar el cervell. En aquesta tècnica s’enregistren els senyals elèctrics que es produeixen en el còrtex humà a través d’elèctrodes col•locats al cap. Aquesta tècnica, però, presenta algunes limitacions a l’hora de realitzar els enregistraments, la principal limitació es coneix com a artefactes, que són senyals indesitjats que es mesclen amb els senyals EEG. L’objectiu d’aquest treball de final de màster és presentar tres nous mètodes de neteja d’artefactes que poden ser aplicats en EEG. Aquests estan basats en l’aplicació de la Multivariate Empirical Mode Decomposition, que és una nova tècnica utilitzada per al processament de senyal. Els mètodes de neteja proposats s’apliquen a dades EEG simulades que contenen artefactes (pestanyeigs), i un cop s’han aplicat els procediments de neteja es comparen amb dades EEG que no tenen pestanyeigs, per comprovar quina millora presenten. Posteriorment, dos dels tres mètodes de neteja proposats s’apliquen sobre dades EEG reals. Les conclusions que s’han extret del treball són que dos dels nous procediments de neteja proposats es poden utilitzar per realitzar el preprocessament de dades reals per eliminar pestanyeigs.

MBIS: multivariate Bayesian image segmentation tool.

We present MBIS (Multivariate Bayesian Image Segmentation tool), a clustering tool based on the mixture of multivariate normal distributions model. MBIS supports multichannel bias field correction based on a B-spline model. A second methodological novelty is the inclusion of graph-cuts optimization for the stationary anisotropic hidden Markov random field model. Along with MBIS, we release an evaluation framework that contains three different experiments on multi-site data. We first validate the accuracy of segmentation and the estimated bias field for each channel. MBIS outperforms a widely used segmentation tool in a cross-comparison evaluation. The second experiment demonstrates the robustness of results on atlas-free segmentation of two image sets from scan-rescan protocols on 21 healthy subjects. Multivariate segmentation is more replicable than the monospectral counterpart on T1-weighted images. Finally, we provide a third experiment to illustrate how MBIS can be used in a large-scale study of tissue volume change with increasing age in 584 healthy subjects. This last result is meaningful as multivariate segmentation performs robustly without the need for prior knowledge.

Multivariate analyses of genotype x environment interaction of popcorn

The objectives of this work were to evaluate the genotype x environment (GxE) interaction for popcorn and to compare two multivariate analyses methods. Nine popcorn cultivars were sown on four dates one month apart during each of the agricultural years 1998/1999 and 1999/2000. The experiments were carried out using randomized block designs, with four replicates. The cv. Zélia contributed the least to the GxE interaction. The cv. Viçosa performed similarly to cv. Rosa-claro. Optimization of GxE was obtained for cv. CMS 42 for a favorable mega-environment, and for cv. CMS 43 for an unfavorable environment. Multivariate analysis supported the results from the method of Eberhart & Russell. The graphic analysis of the Additive Main effects and Multiplicative Interaction (AMMI) model was simple, allowing conclusions to be made about stability, genotypic performance, genetic divergence between cultivars, and the environments that optimize cultivar performance. The graphic analysis of the Genotype main effects and Genotype x Environment interaction (GGE) method added to AMMI information on environmental stratification, defining mega-environments and the cultivars that optimized performance in those mega-environments. Both methods are adequate to explain the genotype x environment interactions.

Univariate versus multivariate modeling of panel data

Panel data can be arranged into a matrix in two ways, called 'long' and 'wide' formats (LFand WF). The two formats suggest two alternative model approaches for analyzing paneldata: (i) univariate regression with varying intercept; and (ii) multivariate regression withlatent variables (a particular case of structural equation model, SEM). The present papercompares the two approaches showing in which circumstances they yield equivalent?insome cases, even numerically equal?results. We show that the univariate approach givesresults equivalent to the multivariate approach when restrictions of time invariance (inthe paper, the TI assumption) are imposed on the parameters of the multivariate model.It is shown that the restrictions implicit in the univariate approach can be assessed bychi-square difference testing of two nested multivariate models. In addition, commontests encountered in the econometric analysis of panel data, such as the Hausman test, areshown to have an equivalent representation as chi-square difference tests. Commonalitiesand differences between the univariate and multivariate approaches are illustrated usingan empirical panel data set of firms' profitability as well as a simulated panel data.

Preoperative nutritional screening by the specialist instead of the nutritional risk score might prevent excess nutrition: a multivariate analysis of nutritional risk factors.

BACKGROUND: The aim of the current study was to assess whether widely used nutritional parameters are correlated with the nutritional risk score (NRS-2002) to identify postoperative morbidity and to evaluate the role of nutritionists in nutritional assessment. METHODS: A randomized trial on preoperative nutritional interventions (NCT00512213) provided the study cohort of 152 patients at nutritional risk (NRS-2002 ≥3) with a comprehensive phenotyping including diverse nutritional parameters (n=17), elaborated by nutritional specialists, and potential demographic and surgical (n=5) confounders. Risk factors for overall, severe (Dindo-Clavien 3-5) and infectious complications were identified by univariate analysis; parameters with P<0.20 were then entered in a multiple logistic regression model. RESULTS: Final analysis included 140 patients with complete datasets. Of these, 61 patients (43.6%) were overweight, and 72 patients (51.4%) experienced at least one complication of any degree of severity. Univariate analysis identified a correlation between few (≤3) active co-morbidities (OR=4.94; 95% CI: 1.47-16.56, p=0.01) and overall complications. Patients screened as being malnourished by nutritional specialists presented less overall complications compared to the not malnourished (OR=0.47; 95% CI: 0.22-0.97, p=0.043). Severe postoperative complications occurred more often in patients with low lean body mass (OR=1.06; 95% CI: 1-1.12, p=0.028). Few (≤3) active co-morbidities (OR=8.8; 95% CI: 1.12-68.99, p=0.008) were related with postoperative infections. Patients screened as being malnourished by nutritional specialists presented less infectious complications (OR=0.28; 95% CI: 0.1-0.78), p=0.014) as compared to the not malnourished. Multivariate analysis identified few co-morbidities (OR=6.33; 95% CI: 1.75-22.84, p=0.005), low weight loss (OR=1.08; 95% CI: 1.02-1.14, p=0.006) and low hemoglobin concentration (OR=2.84; 95% CI: 1.22-6.59, p=0.021) as independent risk factors for overall postoperative complications. Compliance with nutritional supplements (OR=0.37; 95% CI: 0.14-0.97, p=0.041) and supplementation of malnourished patients as assessed by nutritional specialists (OR=0.24; 95% CI: 0.08-0.69, p=0.009) were independently associated with decreased infectious complications. CONCLUSIONS: Nutritional support based upon NRS-2002 screening might result in overnutrition, with potentially deleterious clinical consequences. We emphasize the importance of detailed assessment of the nutritional status by a dedicated specialist before deciding on early nutritional intervention for patients with an initial NRS-2002 score of ≥3.

Role of brachytherapy in the treatment of cancers of the anal canal. Long-term follow-up and multivariate analysis of a large monocentric retrospective series.

BACKGROUND AND PURPOSE: There are few data on long-term clinical results and tolerance of brachytherapy in anal canal cancer. We present one of the largest retrospective analyses of anal canal cancers treated with external beam radiotherapy with/without (±) chemotherapy followed by a brachytherapy boost. MATERIALS AND METHODS: We performed a retrospective analysis of clinical results in terms of efficacy and toxicity. The impact of different clinical and therapeutic variables on these outcomes was studied. RESULTS: From May 1992 to December 2009, 209 patients received brachytherapy after external beam radiotherapy ± chemotherapy. Of these patients, 163 were stage II or stage IIIA (UICC 2002) and 58 were N1-3. According to age, ECOG performance status (PS), and comorbidities, patients received either radiotherapy alone (58/209) or radiochemotherapy (151/209). The median follow-up was 72.8 months. The 5- and 10-year local control rates were 78.6 and 73.9 %, respectively. Globally, severe acute and late G3-4 reactions (NCI-CTC scale v. 4.0) occurred in 11.2 and 6.3 % of patients, respectively. Univariate analysis showed the statistical impact of the pelvic treatment volume (p = 0.046) and of the total dose (p = 0.02) on the risk of severe acute and late toxicities, respectively. Only six patients required permanent colostomy because of severe late anorectal toxicities. CONCLUSION: After a long follow-up time, brachytherapy showed an acceptable toxicity profile and high local control rates in patients with anal canal cancer.

Eliminació d'artefactes en EGG mitjançant l'ús de la Multivariate Empirical Mode Decomposition

La tècnica de l’electroencefalograma (EEG) és una de les tècniques més utilitzades per estudiar el cervell. En aquesta tècnica s’enregistren els senyals elèctrics que es produeixen en el còrtex humà a través d’elèctrodes col•locats al cap. Aquesta tècnica, però, presenta algunes limitacions a l’hora de realitzar els enregistraments, la principal limitació es coneix com a artefactes, que són senyals indesitjats que es mesclen amb els senyals EEG. L’objectiu d’aquest treball de final de màster és presentar tres nous mètodes de neteja d’artefactes que poden ser aplicats en EEG. Aquests estan basats en l’aplicació de la Multivariate Empirical Mode Decomposition, que és una nova tècnica utilitzada per al processament de senyal. Els mètodes de neteja proposats s’apliquen a dades EEG simulades que contenen artefactes (pestanyeigs), i un cop s’han aplicat els procediments de neteja es comparen amb dades EEG que no tenen pestanyeigs, per comprovar quina millora presenten. Posteriorment, dos dels tres mètodes de neteja proposats s’apliquen sobre dades EEG reals. Les conclusions que s’han extret del treball són que dos dels nous procediments de neteja proposats es poden utilitzar per realitzar el preprocessament de dades reals per eliminar pestanyeigs.

Advanced chemometric methods: Applicability on industrial data

Recent years have produced great advances in the instrumentation technology. The amount of available data has been increasing due to the simplicity, speed and accuracy of current spectroscopic instruments. Most of these data are, however, meaningless without a proper analysis. This has been one of the reasons for the overgrowing success of multivariate handling of such data. Industrial data is commonly not designed data; in other words, there is no exact experimental design, but rather the data have been collected as a routine procedure during an industrial process. This makes certain demands on the multivariate modeling, as the selection of samples and variables can have an enormous effect. Common approaches in the modeling of industrial data are PCA (principal component analysis) and PLS (projection to latent structures or partial least squares) but there are also other methods that should be considered. The more advanced methods include multi block modeling and nonlinear modeling. In this thesis it is shown that the results of data analysis vary according to the modeling approach used, thus making the selection of the modeling approach dependent on the purpose of the model. If the model is intended to provide accurate predictions, the approach should be different than in the case where the purpose of modeling is mostly to obtain information about the variables and the process. For industrial applicability it is essential that the methods are robust and sufficiently simple to apply. In this way the methods and the results can be compared and an approach selected that is suitable for the intended purpose. Differences in data analysis methods are compared with data from different fields of industry in this thesis. In the first two papers, the multi block method is considered for data originating from the oil and fertilizer industries. The results are compared to those from PLS and priority PLS. The third paper considers applicability of multivariate models to process control for a reactive crystallization process. In the fourth paper, nonlinear modeling is examined with a data set from the oil industry. The response has a nonlinear relation to the descriptor matrix, and the results are compared between linear modeling, polynomial PLS and nonlinear modeling using nonlinear score vectors.