Search

In this paper least squares and maximum likelihood estimates under univariate and multivariate linear models with a priori information related to maximum effects in the models are determined. Both loss functions (the least squares and negative log-likelihood) and the constraints are convex, so the convex optimization theory can be utilized to obtain estimates, which in this paper are called Safety belt estimates. In particular, the complementary slackness condition, common in convex optimization, implies two alternative types of solutions, strongly dependent on the data and the restriction. It is experimentally shown that, despite of the similarity to the ridge regression estimation under the univariate linear model, the Safety belt estimates behave usually better than estimates obtained via ridge regression. Moreover, concerning the multivariate model, the proposed technique represents a completely novel approach., This is a technical report and has not been externally reviewed.

Quadratic discriminant analysis is a well-established supervised classification method, which extends the linear the linear discriminant analysis by relaxing the assumption of equal variances across classes. In this study, quadratic discriminant analysis is used to develop a quadratic classification rule based on repeated measurements. We employ a bilinear regression model to assign new observations to predefined populations and approximate the misclassification probability. Through weighted estimators, we estimate unknown mean parameters and derive moments of the quadratic classifier. We then conduct numerical simulations to compare misclassification probabilities using true and estimated mean parameters, as well as probabilities computed through simulation. Our findings suggest that as the distance between groups widens, the misclassification probability curve decreases, indicating that classifying observations is easier in widely separated groups compared to closely clustered ones., This report has not been peer reviewed.

The likelihood approach used in this paper leads to quadratic discriminant functions. Classification into one of two known multivariate normal populations with a known and unknown covariance matrix are separately considered, where the two cases depend on the sample size and an unknown squared Mahalanobis distance. Their exact distributions are complicated to obtain. Therefore, moments for the likelihood based discriminant functions are established to express the basic characteristics of respective distribution.

The exact distribution of a classification function is often complicated to allow for easy numerical calculations of misclassification errors. The use of expansions is one way of dealing with this difficulty. In this paper, approximate probabilities of misclassification of the maximum likelihood-based discriminant function are established via an Edgeworth-type expansion based on the standard normal distribution for discriminating between two multivariate normal populations.

In this paper, probabilities of misclassification of a two-step likelihood-based discriminant rule are established for the classification of growth curves. The defined two-step classifier considers the fact that the growth curves might not belong to any of the two predetermined populations. The distribution for the classifier is approximated via an Edgeworth-type expansion.

The possibility that a new observation can be allocated to an unknown population is considered. von Rosen and Singull (2022) derived a classi cation rule taking into account this perspective. The classi cation rule consists of two criteria. In this paper, the mean and variance of these criteria needed to discriminate between two growth curves are established., This is a technical report and has not been externally reviewed.

In this paper we consider discrimination between two populations of repeated measurements using growth curve models. We establish a classification procedure that is likelihood based, in that sense that we compare the two likelihoods given that the new observation belongs to respectively population. We also discuss the possibility that we classify the new observation to an unknown population, which we show is natural when considering growth curves.

The likelihood approach used in this paper leads to quadratic discriminant functions. Classification with a known and unknown covariance matrix are separately considered, where the two cases depend on the sample size and an unknown squared Mahalanobis distance. Their exact distributions are complicated to obtain. Therefore, moments for the likelihood based discriminant functions are established to express the basic characteristics of respective distribution.

Panel survey data from Uganda, as well as data from the 2014 Uganda Population and Housing Census data have been analysed. The Growth Curve model with rank restrictions on parameters was used to estimate the small area means. The aim of the analysis was to assess change over time in household living standards (welfare), i.e., to investigate whether households display growth in living standards? whether households grow at the same rates? and whether households in different geographical areas of the country grow at the same rates? Using a GMANOVA-MANOVA model with rank restrictions on parameters, it was established that growth in household standards of living in Uganda varied across small areas. Sub-regions (small areas) with the highest standards of living in Uganda at the endline were Central Urban region, Kampala Urban region and South Western Urban region, while the sub-regions with the lowest standards of living at the endline were North East Rural region, North East Urban region and Eastern Rural (region). The sub-regions with the highest growth rates in standards of living were Mid West Urban region, Mid North Rural region, and South Western Urban region. The sub-regions with the highest decline in standards of living were East Central Rural region, East Rural region and West Nile Urban region.

The exact distribution of a classification function is often complicated to allow for easy numerical calculations of misclassification errors. The use of expansions is one way of dealing with this diculty. In this paper, approximate probabilities of misclassification of the maximum likelihood based discriminant function are established via an Edgeworth-type expansion based on the standard normal distribution for discriminating between two multivariate normal populations.

Liquid financial markets, such as the options market of the S&P 500 index, create vast amounts of data every day, i.e., so-called intraday data. However, this highly granular data is often reduced to single-time when used to estimate financial quantities. This under-utilization of the data may reduce the quality of the estimates. In this paper, we study the impacts on estimation quality when using intraday data to estimate dividends. The methodology is based on earlier linear regression (ordinary least squares) estimates, which have been adapted to intraday data. Further, the method is also generalized in two aspects. First, the dividends are expressed as present values of future dividends rather than dividend yields. Second, to account for heteroscedasticity, the estimation methodology was formulated as a weighted least squares, where the weights are determined from the market data. This method is compared with a traditional method on out-of-sample S&P 500 European options market data. The results show that estimations based on intraday data have, with statistical significance, a higher quality than the corresponding single-times estimates. Additionally, the two generalizations of the methodology are shown to improve the estimation quality further.

The likelihood approach used in this paper leads to quadratic discriminant functions. Classification into one of two known multivariate normal populations with a known and unknown covariance matrix are separately considered, where the two cases depend on the sample size and an unknown squared Mahalanobis distance. Their exact distributions are complicated to obtain. Therefore, moments for the likelihood based discriminant functions are established to express the basic characteristics of respective distribution.

We propose asymptotic approximations for the probabilities of misclassification in linear discriminant analysis when the group means follow a growth curve structure. The discriminant function can classify a new observation vector of p repeated measurements into one of several multivariate normal populations with equal covariance matrix. We derive certain relations of the statistics under consideration in order to obtain asymptotic approximation of misclassification errors for the two group case. Finally, we perform Monte Carlo simulations to evaluate the reliability of the proposed results., Funding Agencies|Swedish Research CouncilSwedish Research CouncilEuropean Commission [2017-03003]

Let X be a x random matrix whose entries are independent and identically distributed real random variables with zero mean and unit variance. We study the limiting behaviors of the 2-normal condition number k(p,n) of X in terms of large deviations for large n, with p being fixed or with . We propose two main ingredients: (i) to relate the large-deviation probabilities of k(p,n) to those involving n independent and identically distributed random variables, which enables us to consider a quite general distribution of the entries (namely the sub-Gaussian distribution), and (ii) to control, for standard normal entries, the upper tail of k(p,n) using the upper tails of ratios of two independent random variables, which enables us to establish an application in statistical inference.

The biggest challenge of treating HIV is rampant liver-related morbidity and mortality. This is, to some extent, attributed to hepatocytes acting as viral reservoirs to both HIV and HBV. Viral reservoirs harbour latent provirus, rendering it inaccessible by combinational antiretroviral therapy (cART) that is specific to actively proliferating virus. Latency reversal agents (LRA) such as Shock and kill or lock and block, aiming at activating the latently infected cells, have been developed. However, they are CD4+ cell-specific only. There is evidence that the low replication level of HIV in hepatocytes is mainly due to the latency of the provirus in these cells. LRA are developed to reduce the number of latently infected cells; however, the impact of the period viral latency in hepatocytes especially, during HIV/HBV coinfection, needs to be investigated. Viral coinfection coupled with lifelong treatment of HIV/HBV necessitates investigation for the optimal control strategy. We propose a coinfection mathematical model with delay and use optimal control theory to analyse the effect of viral latency in hepatocytes on the dynamics of HIV/HBV coinfection. Analytical results indicate that HBV cannot take a competitive exclusion against HIV; thus, the coinfection endemic equilibrium implies chronic HBV in HIV-infected patients. Numerical and analytical results indicate that both HIV and HBV viral loads are higher with longer viral latency period in hepatocytes, which indicates the need to upgrade LRA to other non-CD4+ cell viral reservoirs. Higher viral load caused by viral latency coupled with the effects of cART partly explains why liver-related complications are the leading cause of mortality in HIV-infected persons., Funding Agencies|Sida Phase 4 Bilateral Program; Makerere University 2015-2020; Kyambogo University; Linkopings University

In this paper, predictions of future price movements of a major American stock index were made by analyzing past movements of the same and other correlated indices. A model that has shown very good results in audio and speech generation was modified to suit the analysis of financial data and was then compared to a base model, restricted by assumptions made for an efficient market. The performance of any model, trained by looking at past observations, is heavily influenced by how the division of the data into train, validation and test sets is made. This is further exaggerated by the temporal structure of the financial data, which means that the causal relationship between the predictors and the response is dependent on time. The complexity of the financial system further increases the struggle to make accurate predictions, but the model suggested here was still able to outperform the naive base model by more than 20% and 37%, respectively, when predicting the next day’s closing price and the next day’s trend.

Obstetric fistula is among the most severe maternal morbidities. It is caused directly by obstructed labour, and indirectly by social, political, and economic factors. Effects of obstetric fistula include chronic urinary incontinence. Urinary incontinence often leads to isolation and death if left untreated. Death of women during childbirth is detrimental to both social economic transformation and shift of poor economies to middle income status. In this study, we estimate the burden of obstetric fistula on Ugandan women, and the effect of educating these women on its causes. We hypothesise that symptomatic women seek treatment after interactions with a treated woman. We consider one cycle of fistula and seek to determine how many women will eventually seek treatment, when one woman is treated in a population. We further assume that each treated woman becomes the educational resource for more symptomatic women. We then invest in getting women into treatment, providing treatment, or educating them about prevention and treatment of obstetric fistula. Results show that for each investment strategy, there exists a unique maximal solution that remains bounded. It is also shown from the model that high investment rates leads to high treatments rates. However, the highest treatment output is obtained when the level of investment is 44.9%. We also see from the results that for every one woman treated, 17 more seek treatment. However, as investment increases, the number of women that are eventually treated reduces due to lack of further investment into treatment and therefore no more women to treat. Results also show that educating women reduced fistula prevalence although the magnitude of reduction depended on the level of education coverage. These results show that an optimal strategy to reduce fistula prevalence would require achieving a balance between the two desirable but incompatible health outcomes. This will help maximise the total return or effectiveness under given, Funding Agencies|SIDA bilateral program Makerere

This volume is a tribute to Professor Dietrich von Rosen on the occasion of his 65th birthday. It contains a collection of twenty original papers. The contents of the papers evolve around multivariate analysis and random matrices with topics such as high-dimensional analysis, goodness-of-fit measures, variable selection and information criteria, inference of covariance structures, the Wishart distribution and growth curve models.