No description available.

Joint models for longitudinal and survival data are a class of models that jointly analyse an outcome repeatedly observed over time such as a bio-marker and associated event times. These models are useful in two practical applications; firstly focusing on survival outcome whilst accounting for time varying covariates measured with error and secondly focusing on the longitudinal outcome while controlling for informative censoring. Interest on the estimation of these joint models has grown in the past two and half decades. However, minimal effort has been directed towards developing diagnostic assessment tools for these models. The available diagnostic tools have mainly been based on separate analysis of residuals for the longitudinal and survival sub-models which could be sub-optimal. In this thesis we make four contributions towards the body of knowledge. We first developed influence diagnostics for the shared parameter joint model for longitudinal and survival data based on Cook's statistics. We evaluated the performance of the diagnostics using simulation studies under different scenarios. We then illustrated these diagnostics using real data set from a multi-center clinical trial on TB pericarditis (IMPI). The second contribution was to implement a variance shift outlier model (VSOM) in the two-stage joint survival model. This was achieved by identifying outlying subjects in the longitudinal sub-model and down-weighting before the second stage of the joint model. The third contribution was to develop influence diagnostics for the multivariate joint model for longitudinal and survival data. In this setting we considered two longitudinal outcomes, square root CD4 cell count which was Gaussian in nature and antiretroviral therapy (ART) uptake which was binary. We achieved this by extending the univariate case i based on Cook's statistics for all parameters. The fourth contribution was to implement influence diagnostics in joint models for longitudinal and survival data with multiple failure types (competing risk). Using IMPI data set we considered two competing events in the joint model; death and constrictive pericarditis. Using simulation studies and IMPI dataset the developed diagnostics identified influential subjects as well as observations. The performance of the diagnostics was over 98% in simulation studies. We further conducted sensitivity analyses to check the impact of influential subjects and/or observations on parameter estimates by excluding them and re-fitting the joint model. We observed subtle differences, overall in the parameter estimates, which gives confidence that the initial inferences are credible and can be relied on. We illustrated case deletion diagnostics using the IMPI trial setting, these diagnostics can also be applied to clinical trials with similar settings. We therefore make a strong recommendation to analysts to conduct influence diagnostics in the joint model for longitudinal and survival data to ascertain the reliability of parameter estimates. We also recommend the implementation of VSOM in the longitudinal part of the two-stage joint model before the second stage.

No description available.

With computers being embedded in every walk of our life, there is an increasing demand forintuitive devices for human-computer interaction. As human beings use gestures as importantmeans of communication, devices based on gesture recognition systems will be effective for humaninteraction with computers. However, it is very important to keep such a system as non-intrusive aspossible, to reduce the limitations of interactions. Designing such non-intrusive, intuitive, camerabasedreal-time gesture recognition system has been an active area of research research in the fieldof computer vision.Gesture recognition invariably involves tracking body parts. We find many research works intracking body parts like eyes, lips, face etc. However, there is relatively little work being done onfull body tracking. Full-body tracking is difficult because it is expensive to model the full-body aseither 2D or 3D model and to track its movements.In this work, we propose a monocular gesture recognition system that focuses on recognizing a setof arm movements commonly used to direct traffic, guiding aircraft landing and for communicationover long distances. This is an attempt towards implementing gesture recognition systems thatrequire full body tracking, for e.g. an automated recognition semaphore flag signaling system.We have implemented a robust full-body tracking system, which forms the backbone of ourgesture analyzer. The tracker makes use of two dimensional link-joint (LJ) model, which representsthe human body, for tracking. Currently, we track the movements of the arms in a video sequence,however we have future plans to make the system real-time. We use distance transform techniquesto track the movements by fitting the parameters of LJ model in every frames of the video captured.The tracker's output is fed a to state-machine which identifies the gestures made. We haveimplemented this system using four sub-systems. Namely1. Background subtraction sub-system, using Gaussian models and median filters.2. Full-body Tracker, using L-J Model APIs3. Quantizer, that converts tracker's output into defined alphabets4. Gesture analyzer, that reads the alphabets into action performed.Currently, our gesture vocabulary contains gestures involving arms moving up and down which canbe used for detecting semaphore, flag signaling system. Also we can detect gestures like clappingand waving of arms.

Multi-state models are useful for modelling progression of a disease, where states represent the health status of a subject under study. In practice, patients may be observed when necessity strikes thus implying that the disease and observation processes are not independent. Often, clinical visits are postponed or advanced by the clinician or patient themselves based on the health status of the patient. In such situations, there is a potential for the frequency and timing of the examinations to be dependent on the latent transition times, and the observation process may be informative. We consider the case where the exact times of transitions between health states of the patient are not observed and so the disease process is interval censored. We model the disease and observation processes jointly to ensure valid inference. The transition intensities are modelled assuming proportional hazards and we link the two processes via random effects. Using a Bayesian framework we apply our joint model to the analysis of a large study examining cancer trajectories of palliative care patients. / Graduate

Ingeniero Civil de Minas / El estudio del comportamiento de roca intacta es el punto de partida para poder realizar análisis a escala de macizo rocoso. Es por esto que se han realizado grandes esfuerzos para replicar y validar dicho comportamiento mediante el modelamiento numérico. Aquí es donde aparece el programa de elementos discretos PFC3D, el cual replica a la roca intacta mediante esferas rígidas unidas entre sí por modelos de contactos, cuyo comportamiento depende de micro-parámetros. El modelo de contactos utilizado últimamente para realizar los ensayos con este programa tiene por nombre Enhanced Bonded Particle Model, el cual presenta una serie de deficiencias. De esta manera se trabajó en la creación de un nuevo modelo llamado Flat Joint Model, con la finalidad de tratar de superar las falencias que presenta el Enhanced BPM. De esto nace el objetivo principal de este trabajo, el cual consiste en calibrar este nuevo modelo de contacto, de manera de representar el comportamiento de la roca intacta, que en este caso corresponde a la roca Westerly granite, y poder así comparar los resultados obtenidos por ambos modelos. Para lograr la calibración del Flat Joint Model, y tener un punto de comparación con datos de laboratorio, se extraen desde la literatura una serie de datos experimentales de diversos autores que realizaron una serie de ensayos de roca a muestras de granito que llevan por nombre Westerly granite. Además se cuenta con los resultados de simulaciones hechas con el modelo Enhanced Bonded Particle Model. Los resultados de las simulaciones hechas con el Flat Joint Model, luego de haber sido calibrado, nos dicen que este modelo replica con mayor exactitud los parámetros elásticos y de resistencia en el ensayo de compresión uniaxial y de tracción directa, como también representa de mejor manera la envolvente de falla experimental de la roca, sobre todo a confinamientos bajo los 60 [MPa]. Además se resuelve en parte una de las deficiencias del modelo Enhanced BPM, el cual no es capaz de replicar los valores de la razón de Poisson de la roca obteniendo resultados muy por debajo a los experimentales (diferencias del orden de 86%). Finalmente se concluye que el FJM es mejor modelo de contactos que el Enhanced BPM debido a que es capaz de replicar un comportamiento post-peak de la roca (comportamiento frágil), lo cual era la gran deficiencia del primer modelo. Si bien no se logra representar la fragilidad de la roca a grandes confinamientos, es un gran avance que se tengan comportamientos frágiles aún a 30 [MPa] de confinamiento, siendo que en el Enhanced BPM, a confinamientos de 5 [MPa] ya se presenta una respuesta dúctil de la curva esfuerzo-deformación.

The rapid progress in high-throughput omics methods and high-resolution morphological profiling, coupled with the significant advances in machine learning (ML) and deep learning (DL), has opened new avenues for tackling the notoriously difficult problem of predicting the Mechanism of Action (MoA) for a drug of clinical interest. Understanding a drug's MoA can enrich our knowledge of its biological activity, shed light on potential side effects, and serve as a predictor of clinical success.  This project aimed to examine whether incorporating gene expression data from LINCS L1000 public repository into a joint model previously developed by Tian et al. (2022), which combined chemical structure and morphological profiles derived from Cell Painting, would have a synergistic effect on the model's ability to classify chemical compounds into ten well-represented MoA classes. To do this, I explored the gene expression dataset to assess its quality, volume, and limitations. I applied a variety of ML and DL methods to identify the optimal single model for MoA classification using gene expression data, with a particular emphasis on transforming tabular data into image data to harness the power of convolutional neural networks. To capitalize on the complementary information stored in different modalities, I tested end-to-end integration and soft-voting on sets of joint models across five stratified data splits.  The gene expression dataset was relatively low in quality, with many uncontrollable factors that complicated MoA prediction. The highest-performing gene expression model was a one-dimensional convolutional neural network, with an average macro F1 score of 0.40877 and a standard deviation of 0.034. Approaches converting tabular data into image data did not significantly outperform other methods. Combining optimized single models resulted in a performance decline compared to the best single model in the combination. To take full advantage of algorithmic developments in drug development and high-throughput multi-omics data, my project underscores the need for standardizing data generation and optimizing data fusion methods.

No description available.

No description available.

No description available.