Departments of Statistics

Permanent URI for this collectionhttps://etd.hu.edu.et/handle/123456789/102

Browse

Start date: 2023End date: 2025

Search Results

Now showing 1 - 10 of 15
  • No Thumbnail Available
    Item
    Spatio-Temporal Modelling Spatially Aggregated Data, In case of Malaria in Southern Ethiopia
    (HAWASSA UNIVERSITY, 2023-10) YONAS SHUKE KITAWA
    Malaria is a major public health concern worldwide, particularly in Sub-Saharan Africa. Ethiopia accounts for 1.7% of cases and 1.5% of global deaths, with seasonal and unstable transmission patterns. The COVID-19 pandemic has increased morbidity, making early de- tection crucial for mitigating the negative effects of malaria. Thus disease risk mapping is important to identify areas with elevated risks to assist police decisions. However, in such areas, the data sets are mostly available in aggregated form. Spatially aggregated data is often expressed as disease cases or average measurements from districts, often focused on administrative convenience rather than knowledge of the aetiology of the disease. When the underlying results in the process of disease outcomes are thought to be spatially con- tinuous, a spatially continuous model should be discovered rather than typically employed spatially discrete models. This thesis is composed of three papers; in the first paper, We have developed an early warning system for malaria that provides a signal whenever the in- cidence of malaria exceeds certain thresholds. Then, we identified areas with some elevated risks based on temporal trends. Our findings could assist public health offices in identifying areas for early intervention and guiding healthcare resource allocation, allowing its limited resources to be utilized to the greatest extent possible. In addition, we explored the sig- nificance of incorporating environmental variables in forecasting malaria risk in Southern Ethiopia. In the second paper, the researcher examined a spatiotemporal model to highlight disease risk change over time. Four models were considered to be the most appropriate for the problem of spatially aggregated data in a small area. According to the findings of the vii study, the spatiotemporal spatially discrete approximation to the log-Gaussian Cox process (SPSDLGCP) gives credible and computationally efficient estimates of disease risk on both spatially continuous and aggregated scales. Furthermore, we have discovered that including spatial correlation is critical when modelling the spatiotemporal variation of disease risk in the region. In the third paper, taking into account areal units as the vertices of a graph and neighbour interactions as a collection of edges, we proposed a graph-based optimization approach that could be used to estimate either a static or a temporally changing neighbour- hood matrix to better capture the spatial correlation in the data set. When compared to the commonly utilized border-sharing rule, the strategy yielded better inference. In particular, using a temporal varying waiting matrix for modelling the aggregated count of two domi- nating Plasmodium species; P. falciparum and P. vivax ; a clearer picture of the distribution of the incidence in the region was provided. We have also highlighted regions where either species poses unusual threats. Finally, we identified major climate variables linked with malaria risk in the region, used a spline function to integrate the non-linear relationship between climatic factors and malaria risk, and investigated the delayed climatic impact on malaria risk. From August 2013 to June 2019, we considered aggregated malaria counts in 149 districts located in Southern Ethiopia. The findings of our study have the potential to assist the SNNPRS public health department and other stakeholders in defining areas for early intervention as well as regulating the distribution of limited resources for the healthcare facility to the greatest extent possible. Furthermore, by looking for a more toiled approach for estimating the waiting matrix for multivariate cases, including additional risk factors that were not identified, risk discontinuities, and the problem of zero inflation, one can improve the existing methods.
  • No Thumbnail Available
    Item
    Extremal Random Forests, Tree-Based Machine Learning Methods and Extreme Value Theory For Vehicle Insurance Data
    (HAWASSA UNIVERSITY, 2023-04) EDOSSA MERGA TEREFE
    Classical methods for quantile regression fail in cases where the quantile of interest is extreme and only few or no training data points exceed it. Asymptotic results from extreme value theory can be used to extrapolate beyond the range of the data, and several approaches exist that use linear regression, kernel methods or generalized additive models. Most of these methods break down if the predictor space has more than a few dimensions or if the regression function of extreme quantiles is complex. We propose a method for extreme quantile regression that combines the flexibility of random forests with the theory of extrapolation. Our extremal random forest (ERF) estimates the parameters of a generalized Pareto distribution, conditional on the predictor vector, by maximizing a local likelihood with weights extracted from a quantile random forest. Under certain assumptions, we show consistency of the estimated parameters. Furthermore, we penalize the shape parameter in this likelihood to regularize its variability in the predictor space. Simulation studies show that our ERF outperforms both classical quantile regression methods and existing regression approaches from extreme value theory. We apply our methodology to extreme quantile prediction for U.S. wage data.
  • No Thumbnail Available
    Item
    APPLICATION OF MULTISTATE MARKOV MODEL ON THE PROGRESSION OF CHRONIC KIDNEY DISEASE PATIENTS AT TIKUR ANBESSA SPECIALIZED HOSPITAL, TASH, ADDIS ABABA, ETHIOPIA.
    (HAWASSA UNIVERSITY, 2023-10) ZELALEM TOLOSA
    Background: Chronic kidney disease (CKD) is a serious issue for public health. According to the WHO report for 2022, 17 million people will die from NCD diabetes (2.0 million including kidney disease deaths caused by diabetes) before the age of 70, with low- and middle-income countries providing for 86% of these premature deaths. 77% of NCD-related deaths occur in low- and middle-income countries including Ethiopia. is one of the low- and middle-income nations in the sub-Saharan region. This study was aimed to estimated the effect of covariates on progression between different stages of CKD among patients under follow up treatment at Tikur Anbessa Specialized Hospital TASH, Addis Ababa, Ethiopia, using Multistate Markov Model. Method: The study was carried out using a retrospective cohort study design on 267 CKD patients age greater than 18 randomly selected at nephrology clinic of TASH who start follow up in May 2018 up to April 2023 for five years. The five stages of CKD disease defined based on the Kidney Disease Improving Global Outcome (KDIGO) guidelines with make only forward transition among different transient stage continuously considered in the Multi-state Markov Model to estimate the transition conditional probabilities, transition intensity rate, total length of stay in different CKD patient’s stage. Result: From the total number of patients included in the study (267 CKD patients), 153 (57.3%) were males and 114 (42.7%) were females. Patients in stages 1, 2.3A, 3B, and 4 had an estimated probability of 94% (0.94), 93% (0.93), 93% (0.93), 96% (0.96), and 98% (0.98) of staying in the same stage, respectively, after one month. Estimated sojourn times for states 1, 2, 3A,3B and 4 were 16.5, 14.5, 15.5, 30.3 and 53.8 months respectively. Conclusion: Prognostic factors like being male, having a history of Diabetes, having a history of Hypertension, and having a history of heart disease were the factors that had a higher risk of progressing to severe stages in CKD patients, and Age, Haemoglobin, and Potassium were positively (or harmfully) associated with the progression of eGFR or CKD stages. Whereas Phosphate, Sodium, and Urea were negatively associated with the progression change of eGFR or CKD stages. The transition probability from a given good stage to the next worse stage increases with time, reaches its optimum (peak) at a time, and starts to decline as time goes on. stage 4 CKD had the longest estimated mean duration, followed by stage 3B, while the expected mean duration of stage 2 CKD was the shortest.
  • No Thumbnail Available
    Item
    SYSTEMATIC REVIEW AND META-ANALYSIS ON RISK FACTORS OF BREAST CANCER AMONG WOMEN IN SUB-SAHARAN AFRICA
    (HAWASSA UNIVERSITY, 2023-11) MUSEFA KEDIR
    In the 21st century, cancer, specifically breast cancer is expected to be a leading cause of mortality, affecting people under 70 in many countries. Among the commonly reported cancer types, female breast cancer is the most frequently occurring cancer globally. It is the second most common cause of death for women in Africa and stands as the most common form of cancer affecting women in Sub-Saharan Africa (SSA). The aim of this finding is to identify breast cancer risk factors in Sub- Saharan Africa by using systematic review and meta-analysis. We conducted a systematic search of international databases including PubMed/Medline, Scopus, Google Scholar, and Google engi ne to collect relevant studies from 2000 to March 2023. The random-effects model in R software was employed to estimate pooled odds ratios at a 5% significance level. Publication bias was assessed using Egger and Funnel plot methods, with adjustments made using the Trim and Fill method. A sensitivity analysis was performed to ensure the model's robustness and stability in data analysis. This study analyzed a total of 24 articles, including 22 case–control, 1 cross-sectional, and 1 cohort study, with a combined participant count of 17,321. The findings revealed several variables associated with an increased risk of breast cancer: like positive family history of breast cancer (OR=1.87, 95% CI 1.58 to 2.21), alcohol consumption (OR=1.47, 95% CI 1.11 to 1.96), postmenopausal status of the patients, (OR=1.36, 95% CI 1.02 to 1.81), lower educational level (vs. university level) (OR=1.36, 95% CI 1.10 to 1.70), and residing in an urban area (OR=0.54, 95% CI 0.31 to 0.95) reduced the risk of breast cancer. variables like ever breastfeeding, early age at menarche, physical exercise, BMI, oral contraceptive use and smoking status had no statistically significant association with breast cancer. Theis systematic review and meta-analysis enhance our understanding of breast cancer risk factors in sub-Saharan Africa. The study identified significant association between educational level, residence, alcohol consumption, family history of breast cancer, and menopausal status, while other variables did not link to breast cancer significantly. Smoking also did not display a statistically significant association. This underscores the importa nce of rising awareness about major risk factors, such as education, residence, alcohol intake, fa mily history, and menopausal status, to inform decision making and encourage preventive action like reducing alcohol consumption and adhering to regular screenig.
  • No Thumbnail Available
    Item
    JOINT MODELING OF A RECURRENT EVENT IN PROSTATE CANCER AND TIME TO TERMINATE OF PATIENTS IN TIKUR ANBESSA SPECIALIZED HOSPITAL
    (HAWASSA UNIVERISTY, 2023-10) BY IBSA NURU
    Background: The prostate gland, an organ found in the male reproductive system, is where prostate cancer typically originates. Prostate cancer ranks fourth among malignancies diagnosed in men behind lung, colorectal, and stomach cancers worldwide. Recurring occurrences of the same or different kinds of events for specific people or units across time are referred to as recurrent events such as prostate cancer which is an important clinical indicator and the leading cause of prostate cancer mortality. The major aim of this study was to investigating predictors of prostate cancer recurrence and terminal events (death) due to prostate recurrence. Methodology: to reach the aim, 222 prostate cancer patients, between the study period January 1, 2018 to January 30, 2021, who were registered with detailed, comprehensive personal and medical information were include. The retrospective longitudinal study design was applied and the data were analyzed using joint frailty model on prostate CA and death that are recorded in the oncology department of Tikur Anbessa specialized hospital. To investigate determining factors of prostate cancer recurrence and death, a joint frailty model of the recurrent event and terminal event proposed by Liu and others i.e., the joint frailty proportional hazards model was used alongside reduced models for prostate CA recurrence and terminal event (death). Result: From the total of 432 recurrent observation, about 210 (61.1%) of them experienced recurrence of prostate cancer, 192 of the experienced death (terminated) event and 222 (38.9%) were censored. The shared gamma joint frailty model was chosen as the best fit for the prostate cancer data set based on the value of Likelihood cross validation criterion. From the result of shared gamma joint frailty model smoking, stage of prostate cancer, distance metastasize and Gleason score were significantly associated with recurrence of prostate cancer and death. Conclusion and recommendation: The result of shared gamma joint frailty model shows that the stage (III, IV), smoking, distance metastasize (metastasized tumor) and Gleason score were significantly increases the risk of recurrence of prostate cancer and death. While, quitting smoking may improve patients overall prognosis. Timely detection and management of metastasis in prostate cancer patients are crucial, necessitating focused treatment and surveillance. It is recommended that policy maker, ministry of health and Tikur Anbessa Specialized Hospital are expected to make intervention to improve the management and care of prostate cancer patients, ultimately enhancing their quality of life and prognosis.
  • No Thumbnail Available
    Item
    BAYESIAN APPROACH FOR JOINT MODELING OF TIME TO SEIZURE FREEDOM AND SEIZURE FREQUENCY COUNT IN EPILEPTIC PATIENT AT HAWASSA UNIVERSITY COMPREHENSIVE SPECIALIZED HOSPITAL
    (HAWASSA UNIVERSITY, 2023-11) YITAGESU ESHETU
    Background: Epilepsy was classified as a chronic, non-communicable brain condition by the World Health Organization. Over 50 million people globally have been impacted by epilepsy, one of the most common neurological disorders. In Ethiopia, epilepsy is one of the top 20 killers, and 5.2 out of every 1000 people will suffer from it at some point in their lifetime. The main objective of the study was to investigate predictors of seizure attacks progression and time to seizure freedom among epileptic patients using separate and joint analysis in Bayesian approach. Methodology: The study analyzed data from 203 epileptic patients who initiated anti-epileptic drugs (AEDs) at Hawassa University Comprehensive Specialized Hospital Neurologic Clinic between 1st May 2018 up to 1st May 2023. A retrospective cohort study design was carried out and epileptic patients age greater than 18 years old were used as source of population for this study and also the data obtained from HUCSH. A Bayesian approach for joint modeling is used to analysis time to seizure freedom and seizure frequency count. Results: Out of these patients, 80.3% (163) achieved seizure freedom, while 19.7% (40) were censored due to not achieving seizure freedom within the study period. Analyzing factors influencing seizure outcomes, the study found that Phenytoin usage showed a statistically significant positive effect on seizure reduction, while Phenobarbitone and Sodium Valproate did not exhibit significant effects. Having more treatment sessions had a significant positive effect on reducing seizures. Patients with a partial seizure type showed a significant increase in seizure frequency, while those who exercised, had a family history of epilepsy, or consumed alcohol experienced a significant reduction in seizure frequency. Patients without chronic diseases had significantly fewer seizures. Moreover, patients with co-morbidities or a history of alcohol consumption had a higher frequency of seizures. Conclusion: Bayesian joint modeling revealed that the Weibull survival model and Negative Binomial Zero-Inflated model provided the best fit for survival and count data, respectively. This study's findings contribute to a comprehensive understanding of the factors influencing seizure freedom and seizure frequency in epileptic patients, offering valuable insights for clinical management and treatment strategies.
  • No Thumbnail Available
    Item
    APPLICATION OF MULTISTATE MARKOV MODEL IN ANALYZING THE TRANSI TION OF HYPERTENSION AT HAWASSA UNIVERSITY COMPREHENSIVE SPECI ALIZED HOSPITAL, ETHIOPIA
    (HAWASSA UNIVERSITY, 2023-11) TAYESEW SHEGAW MOLTOT
    Background: Hypertension is the primary cause of cardiovascular death as well as other serious conditions like heart attacks, strokes, chronic heart failure, and kidney failure. In Ethiopia, cardiovascular disease is by far the most common NCD-related cause of death, and hypertension is one of the main risk factors. this study aimed to model hypertension progression and identify factors determining the transition rate between different stages of hypertension among hypertensive patients under follow up at HU-CSH recorded between September 2017 to August 2022 using multistate Markov model by the European Society of Cardiology and the European S ociety of Hypertension's guidelines for blood pressure. Method: Data for this study was obtained from Hawassa university comprehensive specialized Hospital, with a total of 210 hypertensive patients who were under follow up from September 2017 to August 2022 were included in the study. A twenty-four-month transition probability between prehypertension progression stages or state 1 (systolic <140 mm Hg & diast olic <90 mm Hg), grade 1 hypertension or state 2 (systolic 140-159 mm Hg & diastolic 90-99 mm Hg), grade 2 hypertension or state 3 (systolic 160-179 mm Hg diastolic 100-109 mm Hg) and grade 3 or state 4 hypertension (systolic ≥180 mm Hg & diastolic ≥110 mm Hg) and factors determining the rate of progression among patients was estimated. Result and Discussion: Among the total number of patients included in the study, 56.7% were female and 43.3% were male. Among the total number of patients included in the study, 56.7% were female and 43.3% were male. Among them, 22.9% of patients were in state 1 at initial checkup, 36.2 in state 2, 23.8% in state 3 and 17.1% in state 4 hypertension. The estimated 24 m onths transition probability for the hypertensive patients was 15.1 % (95% CI: 0.108, 0.210) from state 1 to state 2, 1.4% (95% CI: 0.009, 0.023) from state 1 hypertension to state 3, 0.2 %(95% CI:0.001, 0.003) from state 1 to state 4, 12.8% (95% CI:0.091, 0.177) from state 2 hypertension to state 3, 2.6% (95% CI:0.017, 0.041) from state 2 hypertension to state 4, 20.5% (95%CI:0.137, 0.290) from state 3 to state 4 and 83.3% (95%CI: 0.768, 0.881), 72.6% (95%CI: 0.660, 0.778), 45.1% (95%CI: 0.368, 0.529) and 67.8% (95%CI: 0.577, 0.777) were the estimate probability of remaining at state 1, state 2, state 3 and state 4 respectively. The mean time a pati ent takes to transition from state to state was estimated and state 1 hypertension had the longest estimated time followed by state 2, while state 3 had the shortest estimated sojourn time. By comparing a likelihood ratio test statistic, the full model fits significantly better than the null model. The observed and expected plots does not have much deviations and assumption of homogeneity of transition rate through the specified time are satisfied. Conclusion: The conditional probability of hypertensive patients from good states to the next worst state are decreasing over time except the first state of hypertension. Being male, older aged, living in urban, taking medication and treatments in the past, history of diabetes (between all states) had high risk of transition from state2 and state 3, state 3and state 4. Being female, younger age, living in rural, and not taking medication in the past between state 1 and state 2 were high risk of hypertension. Having family history of hypertension between state 1 to state 2, state 2 to state 3 and who were not family history of hypertension in state 3 to state 4 had high risk of transition of hypertension.V
  • No Thumbnail Available
    Item
    SURVIVAL ANALYSIS OF TIME TO RECOVERY OF ADIMTTED COVID-19 PATIENTS: IN HAWASSA UNVERCITY REFERAL AND COMPREHENSIV HOSPITAL TEREATMENT CENTER.
    (HAWASSA UNIVERSITY, 2023-06) AMANUEL MERDIKYOS NANA
    Corona virus is one of the major pathogens that primarily target the human respiratory system, which started in Wuhan, China in December 2019, has emerged as a global health and economic security threat with an overwhelming growing incidence worldwide. When the World Health Organization (WHO) declared the disease a global public health emergency, different stakeholders stepped up efforts to convince the world that the disease is a serious problem that needs strong containment measures. The main objective of the study is to identify the determinant risk factors for the recovery of corona virus(covid-19) patients. A study population of 826 total Covid-19 Patients that had been treated at Hawassa University Comprehensive and Referral hospital from September 20, 2013 to January 20, 2014 E.C was included in the study. Descriptive statistics and Kaplan-Meier survival curves were used to estimate and compare the recovery time of corona virus (covid-19) patients among different categorical characteristics of the patients. We used survival time model to analyze the data. The Weibull regression model better fits the recovery time of corona virus (covid-19) than the exponential, log-logistic model and log-normal model. The result showed that out of a total of 826 corona virus (covid-19) patients considered total recovery are 637(77.12%) recovered from covid-19. From the result severity (HR=0.932, p value=0.014), Co-morbidty(HR=0.89,p-value=0.038), other pains out of covid-19(HR=0.7918, P-value=0.006), shortness of breath (HR=0.83,p-value=0.025), severe headache (HR=0.843, p value=0.034) and Age (HR=0.8948, p-value=0.000) were the significant factors for the corona virus(covid-19) patients using Weibull regression model. The model showed that the major factors that affect the recovery time of corona-virus (covid-19) and see the associations factors among patients. Patient’s comorbidities have a major impact on CVID-19; So, health profession should close follow up is required for client admitted with comorbidity and create great awareness about the risk factors the corona virus (covid-19).
  • No Thumbnail Available
    Item
  • No Thumbnail Available
    Item
    APPLICATION OF TIME SERIES ANALYSIS FOR MODELLING AND PREDICTING MONTHLY AVERAGE TEMPERATURE IN GLOBAL AND EAST AFRICA
    (HAWASSA UNIVERSITY, 2024-05) NEBIYU MOHAMMED WOYESO
    Global warming has garnered significant attention in recent years due to its profound implications for the environment, economy, and society. This study aims to model and forecast the monthly average temperature globally and in East Africa using time series analysis. Secondary data from the Berkeley method, spanning from 1850 to the present, were utilized. The average temperatures recorded were 14.103°C globally, 23.025°C in Ethiopia, 23.979°C in Kenya, and 23.019°C in Uganda. After applying the first seasonal differencing to achieve stationarity, statistical models such as SARIMA were employed. Based on Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC), SARIMA(4,0,0)(0,1,1,12) for global temperatures, SARIMA(2,0,2)(3,1,1,12) for Ethiopia, SARIMA(2,0,3)(0,1,1) for Kenya, and SARIMA(1,0,4)(0,1,1,12) for Uganda were identified as the best-fit models. Forecasts for the next two decades indicate an increasing temperature trend across all regions. Model performance was evaluated using metrics such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE), all of which indicated good forecasting accuracy. To capture temperature volatility, ARCH/GARCH models, including ARCH, GARCH, and EGARCH, were applied. The EGARCH models were found to be most effective, with EGARCH(4,1) for global temperatures, EGARCH(3,1) for Ethiopia, EGARCH(3,2) for Kenya, and EGARCH(4,1) for Uganda showing superior performance based on lower AIC and BIC values and higher log-likelihood values. The findings have significant implications for climate adaptation planning, particularly in regions like East Africa that are highly susceptible to climate change impacts. By providing reliable forecasting tools, this study supports efforts in agricultural planning, water resource management, energy demand forecasting, public health, and environmental impact assessment.