Fitting of Mixtures with Unspecified Number of Components Using Cross Validation Distance Estimate

Estimation of the number of mixture components (k) is an unsolved problem. Available methods for estimation of k include bootstrapping the likelihood ratio test statistics and optimizing a variety of validity functionals such as AIC, BIC/MDL, and ICOMP. We investigate the minimization of distance between fitted mixture model and the true density as a method for estimating k. The distances considered are Kullback-Leibler (KL) and “L sub 2”. We estimate these distances using cross validation. A reliable estimate of k is obtained by voting of B estimates of k corresponding to B cross validation estimates of distance. This estimation methods with KL distance is very similar to Monte Carlo cross validated likelihood methods discussed by Smyth (2000). With focus on univariate normal mixtures, we present simulation studies that compare the cross validated distance method with AIC, BIC/MDL, and ICOMP. We also apply the cross validation estimate of distance approach along with AIC, BIC/MDL and ICOMP approach, to data from an osteoporosis drug trial in order to find groups that differentially respond to treatment.

Outcome-based adaptive randomization for binary data in clinical trials---Compare and contrast

Monte Carlo simulation has been conducted to investigate parameter estimation and hypothesis testing in some well known adaptive randomization procedures. The four urn models studied are Randomized Play-the-Winner (RPW), Randomized Pôlya Urn (RPU), Birth and Death Urn with Immigration (BDUI), and Drop-the-Loses Urn (DL). Two sequential estimation methods, the sequential maximum likelihood estimation (SMLE) and the doubly adaptive biased coin design (DABC), are simulated at three optimal allocation targets that minimize the expected number of failures under the assumption of constant variance of simple difference (RSIHR), relative risk (ORR), and odds ratio (OOR) respectively. Log likelihood ratio test and three Wald-type tests (simple difference, log of relative risk, log of odds ratio) are compared in different adaptive procedures. ^ Simulation results indicates that although RPW is slightly better in assigning more patients to the superior treatment, the DL method is considerably less variable and the test statistics have better normality. When compared with SMLE, DABC has slightly higher overall response rate with lower variance, but has larger bias and variance in parameter estimation. Additionally, the test statistics in SMLE have better normality and lower type I error rate, and the power of hypothesis testing is more comparable with the equal randomization. Usually, RSIHR has the highest power among the 3 optimal allocation ratios. However, the ORR allocation has better power and lower type I error rate when the log of relative risk is the test statistics. The number of expected failures in ORR is smaller than RSIHR. It is also shown that the simple difference of response rates has the worst normality among all 4 test statistics. The power of hypothesis test is always inflated when simple difference is used. On the other hand, the normality of the log likelihood ratio test statistics is robust against the change of adaptive randomization procedures. ^

Statistical Hurdle Models for Single Cell Gene Expression: Differential Expression and Graphical Modeling

Thesis (Ph.D.)--University of Washington, 2016-06

On a resampling approach for tests on the number of clusters with mixture model-based clustering of tissue samples

We consider the problem of assessing the number of clusters in a limited number of tissue samples containing gene expressions for possibly several thousands of genes. It is proposed to use a normal mixture model-based approach to the clustering of the tissue samples. One advantage of this approach is that the question on the number of clusters in the data can be formulated in terms of a test on the smallest number of components in the mixture model compatible with the data. This test can be carried out on the basis of the likelihood ratio test statistic, using resampling to assess its null distribution. The effectiveness of this approach is demonstrated on simulated data and on some microarray datasets, as considered previously in the bioinformatics literature. (C) 2004 Elsevier Inc. All rights reserved.

Detecting finite bandwidth periodic signals in stationary noise using the signal coherence spectrum

All signals that appear to be periodic have some sort of variability from period to period regardless of how stable they appear to be in a data plot. A true sinusoidal time series is a deterministic function of time that never changes and thus has zero bandwidth around the sinusoid's frequency. A zero bandwidth is impossible in nature since all signals have some intrinsic variability over time. Deterministic sinusoids are used to model cycles as a mathematical convenience. Hinich [IEEE J. Oceanic Eng. 25 (2) (2000) 256-261] introduced a parametric statistical model, called the randomly modulated periodicity (RMP) that allows one to capture the intrinsic variability of a cycle. As with a deterministic periodic signal the RMP can have a number of harmonics. The likelihood ratio test for this model when the amplitudes and phases are known is given in [M.J. Hinich, Signal Processing 83 (2003) 1349-13521. A method for detecting a RMP whose amplitudes and phases are unknown random process plus a stationary noise process is addressed in this paper. The only assumption on the additive noise is that it has finite dependence and finite moments. Using simulations based on a simple RMP model we show a case where the new method can detect the signal when the signal is not detectable in a standard waterfall spectrograrn display. (c) 2005 Elsevier B.V. All rights reserved.

Avaliação e correções de testes de hipóteses em modelos de sobrevivência com fração de cura

Survival models deals with the modelling of time to event data. In certain situations, a share of the population can no longer be subjected to the event occurrence. In this context, the cure fraction models emerged. Among the models that incorporate a fraction of cured one of the most known is the promotion time model. In the present study we discuss hypothesis testing in the promotion time model with Weibull distribution for the failure times of susceptible individuals. Hypothesis testing in this model may be performed based on likelihood ratio, gradient, score or Wald statistics. The critical values are obtained from asymptotic approximations, which may result in size distortions in nite sample sizes. This study proposes bootstrap corrections to the aforementioned tests and Bartlett bootstrap to the likelihood ratio statistic in Weibull promotion time model. Using Monte Carlo simulations we compared the nite sample performances of the proposed corrections in contrast with the usual tests. The numerical evidence favors the proposed corrected tests. At the end of the work an empirical application is presented.

Finite Population Inference for Causal Parameters

Thesis (Ph.D.)--University of Washington, 2016-08

Power Study on Testing Epidemic Alternatives

Detecting change points in epidemic models has been studied by many scholars. Yao (1993) summarized five existing test statistics in the literature. Out of those test statistics, it was observed that the likelihood ratio statistic showed its standout power. However, all of the existing test statistics are based on an assumption that population variance is known, which is an unrealistic assumption in practice. To avoid assuming known population variance, a new test statistic for detecting epidemic models is studied in this thesis. The new test statistic is a parameter-free test statistic which is more powerful compared to the existing test statistics. Different sample sizes and lengths of epidemic durations are used for the power comparison purpose. Monte Carlo simulation is used to find the critical values of the new test statistic and to perform the power comparison. Based on the Monte Carlo simulation result, it can be concluded that the sample size and the length of the duration have some effect on the power of the tests. It can also be observed that the new test statistic studied in this thesis has higher power than the existing test statistics do in all of cases.

The role of heart rate as a risk marker for predicting adverse outcomes

Cardiovascular disease is one of the leading causes of death around the world. Resting heart rate has been shown to be a strong and independent risk marker for adverse cardiovascular events and mortality, and yet its role as a predictor of risk is somewhat overlooked in clinical practice. With the aim of highlighting its prognostic value, the role of resting heart rate as a risk marker for death and other adverse outcomes was further examined in a number of different patient populations. A systematic review of studies that previously assessed the prognostic value of resting heart rate for mortality and other adverse cardiovascular outcomes was presented. New analyses of nine clinical trials were carried out. Both the original and extended Cox model that allows for analysis of time-dependent covariates were used to evaluate and compare the predictive value of baseline and time-updated heart rate measurements for adverse outcomes in the CAPRICORN, EUROPA, PROSPER, PERFORM, BEAUTIFUL and SHIFT populations. Pooled individual patient meta-analyses of the CAPRICORN, EPHESUS, OPTIMAAL and VALIANT trials, and the BEAUTIFUL and SHIFT trials, were also performed. The discrimination and calibration of the models applied were evaluated using Harrell’s C-statistic and likelihood ratio tests, respectively. Finally, following on from the systematic review, meta-analyses of the relation between baseline and time-updated heart rate, and the risk of death from any cause and from cardiovascular causes, were conducted. Both elevated baseline and time-updated resting heart rates were found to be associated with an increase in the risk of mortality and other adverse cardiovascular events in all of the populations analysed. In some cases, elevated time-updated heart rate was associated with risk of events where baseline heart rate was not. Time-updated heart rate also contributed additional information about the risk of certain events despite knowledge of baseline heart rate or previous heart rate measurements. The addition of resting heart rate to the models where resting heart rate was found to be associated with risk of outcome improved both discrimination and calibration, and in general, the models including time-updated heart rate along with baseline or the previous heart rate measurement had the highest and similar C-statistics, and thus the greatest discriminative ability. The meta-analyses demonstrated that a 5bpm higher baseline heart rate was associated with a 7.9% and an 8.0% increase in the risk of all-cause and cardiovascular death, respectively (both p less than 0.001). Additionally, a 5bpm higher time-updated heart rate (adjusted for baseline heart rate in eight of the ten studies included in the analyses) was associated with a 12.8% (p less than 0.001) and a 10.9% (p less than 0.001) increase in the risk of all-cause and cardiovascular death, respectively. These findings may motivate health care professionals to routinely assess resting heart rate in order to identify individuals at a higher risk of adverse events. The fact that the addition of time-updated resting heart rate improved the discrimination and calibration of models for certain outcomes, even if only modestly, strengthens the case that it be added to traditional risk models. The findings, however, are of particular importance, and have greater implications for the clinical management of patients with pre-existing disease. An elevated, or increasing heart rate over time could be used as a tool, potentially alongside other established risk scores, to help doctors identify patient deterioration or those at higher risk, who might benefit from more intensive monitoring or treatment re-evaluation. Further exploration of the role of continuous recording of resting heart rate, say, when patients are at home, would be informative. In addition, investigation into the cost-effectiveness and optimal frequency of resting heart rate measurement is required. One of the most vital areas for future research is the definition of an objective cut-off value for the definition of a high resting heart rate.

Assessing the fit of unidimensional IRT models for binary data under model misspecification

Model misspecification affects the classical test statistics used to assess the fit of the Item Response Theory (IRT) models. Robust tests have been derived under model misspecification, as the Generalized Lagrange Multiplier and Hausman tests, but their use has not been largely explored in the IRT framework. In the first part of the thesis, we introduce the Generalized Lagrange Multiplier test to detect differential item response functioning in IRT models for binary data under model misspecification. By means of a simulation study and a real data analysis, we compare its performance with the classical Lagrange Multiplier test, computed using the Hessian and the cross-product matrix, and the Generalized Jackknife Score test. The power of these tests is computed empirically and asymptotically. The misspecifications considered are local dependence among items and non-normal distribution of the latent variable. The results highlight that, under mild model misspecification, all tests have good performance while, under strong model misspecification, the performance of the tests deteriorates. None of the tests considered show an overall superior performance than the others. In the second part of the thesis, we extend the Generalized Hausman test to detect non-normality of the latent variable distribution. To build the test, we consider a seminonparametric-IRT model, that assumes a more flexible latent variable distribution. By means of a simulation study and two real applications, we compare the performance of the Generalized Hausman test with the M2 limited information goodness-of-fit test and the Likelihood-Ratio test. Additionally, the information criteria are computed. The Generalized Hausman test has a better performance than the Likelihood-Ratio test in terms of Type I error rates and the M2 test in terms of power. The performance of the Generalized Hausman test and the information criteria deteriorates when the sample size is small and with a few items.

[impact Variables On The Decline In Infant Mortality In The State Of São Paulo, Brazil: 1998-2008].

This is an ecological, analytical and retrospective study comprising the 645 municipalities in the State of São Paulo, the scope of which was to determine the relationship between socioeconomic, demographic variables and the model of care in relation to infant mortality rates in the period from 1998 to 2008. The ratio of average annual change for each indicator per stratum coverage was calculated. Infant mortality was analyzed according to the model for repeated measures over time, adjusted for the following correction variables: the city's population, proportion of Family Health Programs (PSFs) deployed, proportion of Growth Acceleration Programs (PACs) deployed, per capita GDP and SPSRI (São Paulo social responsibility index). The analysis was performed by generalized linear models, considering the gamma distribution. Multiple comparisons were performed with the likelihood ratio with chi-square approximate distribution, considering a significance level of 5%. There was a decrease in infant mortality over the years (p < 0.05), with no significant difference from 2004 to 2008 (p > 0.05). The proportion of PSFs deployed (p < 0.0001) and per capita GDP (p < 0.0001) were significant in the model. The decline of infant mortality in this period was influenced by the growth of per capita GDP and PSFs.

Clinicopathological And Immunohistochemical Evaluation Of Oral And Oropharyngeal Squamous Cell Carcinoma In Chilean Population.

In oral and oropharyngeal squamous cell carcinoma (OCSCC and OPSCC) exist an association between clinical and histopathological parameters with cell proliferation, basal lamina, connective tissue degradation and surrounding stroma markers. We evaluated these associations in Chilean patients. A convenience sample of 37 cases of OCSCC (n=16) and OPSCC (n=21) was analyzed clinically (TNM, clinical stage) and histologically (WHO grade of differentiation, pattern of tumor invasion). We assessed the expression of p53, Ki67, HOXA1, HOXB7, type IV collagen (ColIV) and carcinoma-associated fibroblast (α-SMA-positive cells). Additionally we conducted a univariate/bivariate analysis to assess the relationship of these variables with survival rates. Males were mostly affected (56.2% OCSCC, 76.2% OPSCC). Patients were mainly diagnosed at III/IV clinical stages (68.8% OCSCC, 90.5% OPSCC) with a predominantly infiltrative pattern invasion (62.9% OCSCC, 57.1% OPSCC). Significant association between regional lymph nodes (N) and clinical stage with OCSCC-HOXB7 expression (Chi-Square test P < 0.05) was observed. In OPSCC a statistically significant association exists between p53, Ki67 with gender (Chi-Square test P < 0.05). In OCSCC and OPSCC was statistically significant association between ki67 with HOXA1, HOXB7, and between these last two antigens (Pearson's Correlation test P < 0.05). Furthermore OPSCC-p53 showed significant correlation when it was compared with α-SMA (Kendall's Tau-c test P < 0.05). Only OCSCC-pattern invasion and OPSCC-primary tumor (T) pattern resulted associated with survival at the end of the follow up period (Chi-Square Likelihood Ratio, P < 0.05). Clinical, histological and immunohistochemical features are similar to seen in other countries. Cancer proliferation markers were associated strongly from each other. Our sample highlights prognostic value of T and pattern of invasion, but the conclusions may be limited and should be considered with caution (small sample). Many cases were diagnosed in the advanced stages of the disease, which suggests that the diagnosis of OCSCC and OPSCC is made late.

Association Between Unilateral Tonsillar Enlargement And Lymphoma In Children: A Systematic Review And Meta-analysis.

Lymphoma is the most common head and neck malignancy in children, and palatine tonsils asymmetry is the most frequent clinical manifestation of tonsillar lymphoma. However, several studies with children with tonsillar asymmetry found no case of lymphoma, showing that the relationship of tonsillar asymmetry with lymphoma is unclear. In this review, we aimed to identify the association between tonsillar asymmetry and tonsillar lymphoma in children by conducting systematic reviews of the literature on children with palatine tonsil lymphoma and tonsillar asymmetry. Articles comprising the paediatric age group (up to 18 years) with information concerning clinical manifestations of tonsillar lymphoma or the diagnosis of the tonsillar asymmetry were included. The main cause of asymmetry of palatine tonsils was lymphoid hyperplasia, followed by lymphoma and nonspecific benign changes. The asymmetry of tonsils was present in 73.2% of cases of lymphoma. There was an association between asymmetric palatine tonsils and lymphoma, with a likelihood ratio of 43.5 for children with asymmetry of palatine tonsils and 8938.4 for children with asymmetry of tonsils and other signs of suspicion for malignancy. We also provide recommendations on the management of suspicious cases of palatine tonsil lymphoma.

Inclusão da epistasia em modelo de avaliação genética de bovinos de corte compostos

Dados de bovinos compostos foram analisados para avaliar o efeito da epistasia nos modelos de avaliação genética. As características analisadas foram os pesos aos 205 (P205) e 390 dias (P390) e perímetro escrotal aos 390 dias (PE390). As análises foram realizadas pela metodologia de máxima verossimilhança considerando-se dois modelos: o modelo 1 incluiu como covariáveis os efeitos aditivos diretos e maternos, e os não aditivos das heterozigoses para os efeitos diretos e para o materno total, e o modelo 2 considerou também o efeito direto de epistasia. Para comparação dos modelos, foram utilizados o critério de informação de Akaike (AIC) e o critério de informação Bayesiano de Schwartz (BIC), e o teste de razão de verossimilhança. A inclusão da epistasia no modelo de avaliação genética pouco alterou as estimativas de componentes de (co)variâncias genéticas aditivas e, consequentemente, as herdabilidades. O teste de verossimilhança e o critério de Akaike sugeriram que o modelo 2, que inclui a epistasia, apresentou maior aderência aos dados para todas as características analisadas. O critério BIC indicou este modelo como o melhor apenas para P205. Para análise genética dessa população, o modelo que considerou o efeito de epistasia foi o mais adequado.