Browsing by Subject "15-49 years"
  • Item
    Evaluation of continuation ratio and ordered logistic models in modeling level of educational attainment and its associated factors among females (15-49 years) in Uganda
    (Makerere University, 2025-03) Muwanguzi, Alex. Craig
    Formal education is crucial for individual and economic growth. This study examined the accuracy of the continuation ratio model (CRM) and ordered logistic regression (OLR) in modelling correlates of educational attainment among females aged 15-49 in Uganda. Data from the 2016 Uganda Demographic and Health Survey (UDHS) was used. Descriptive characteristics of respondents were generated. A stepwise ordered logistic regression model was fitted to assess the association between covariates and the outcome, identifying predictor variables for further analysis. At the inferential level, the CRM model was found to be more precise in identifying factors associated with educational attainment. Higher paternal education, joint parental decision-making, wealth status, professional or manual employment, and residence in North Buganda, Bukedi, and Bugisu regions were associated with increased odds of attaining higher education. Conversely, older age, rural residence, larger family size, and residence in Karamoja, Lango, West Nile, Bunyoro, Tooro, and Kigezi regions were associated with lower odds. This study found that the CRM best models female educational attainment, driven by socio-demographic and parental factors. The recommendation is to promote joint parental decision-making, increase economic support for poorer families, and expand employment opportunities for women’s education. Address regional disparities and bridge rural-urban education gaps. We further recommend using this model (CRM) for modelling ordinal response variables representing progress through lower to higher categories.
  • Item
    Factors associated with age at first forced sexual act among women aged 15-49 years in Uganda
    (Makerere University, 2022-11) Akampurira, Innocent
    The main objective of this study was to identify the factors associated with Age at First Forced Sexual Act (FFSA) among women aged 15-49 years in Uganda. Data from 1665 girls and women from the 2016 Uganda Demographic and Health Survey (UDHS) was considered for this study. This sample was obtained from the UDHS data based on the list-wise deletion method to deal with missing data. The assumption that age at first forced sexual act initially increases and then decreases was made for this study and thus a parametric approach was employed. A lognormal distribution was thus assumed for this data and in turn, a log-normal regression model was fitted to examine the factors associated with age at first forced sexual act. The median age at first forced sexual act was 20 years. The highest prevalence (28.5%) of a first forced sexual act was found among girls and women in the region of South Buganda. At bivariable analysis, the log-rank test to check for equality of the survival curves showed that the significant factors associated with FFSA were; education level (χ2(3) = 15.16, p =0.002), wealth index (χ2(4) = 22.67, p < 0.001), region (χ2(14) = 47.53, p < 0.001), employment status (χ2(1) = 4.52, p = 0.034), and age group (χ2(3) = 427.78, p < 0.001). The year of birth was significantly associated with age at first forced sexual act among women in Uganda (p < 0.05). The time to the first forced sexual act among women aged 25-34, 35- 44, and above 44 years in Uganda was 15.6% (TR = 1.156; p<0.001), 26.2% (TR = 1.262; p<0.001) and 40.0% (TR = 1.400; p<0.001) respectively longer than that among those aged 15-24 years to show that older women were least likely to be forced into sex early. Other significant factors associated with age at FFSA were education level, region, religion, and the number of sexual partners including spouse, in the past 12 months. This study highlights that the time of the first experience of forced sex is driven by both socioeconomic and demographic and behavioral factors and therefore efforts directed towards fighting this human rights violation by the Ugandan government and other key stakeholders should consider this aspect.
  • Item
    Modelling haemoglobin among pregnant women as a continuous variable: are the conclusions on its factors influenced by dichotomizing it?
    (Makerere University, 2023-12) Nakitende, Lydia. Mukiibi
    Risk factors for anemia are commonly identified with blood haemoglobin measured and modeled on a binary than continuous scale. This approach leads to information loss and reduced statistical efficiency. In this study, we modeled blood haemoglobin as a continuous response variable that is normally distributed and examined whether inferences on its factors among pregnant women (15-49 years) in Uganda are influenced by dichotomizing it. Data were sourced from the 2016 Uganda Demographic and Health Survey (UDHS). Descriptive statistics and data visualization helped to examine the distribution of blood haemoglobin, and descriptive characteristics of the pregnant women were generated. This was followed by fitting a simple linear regression model to assess the association between each covariate and blood haemoglobin and, in turn, identify explanatory variables for further analysis at the multivariable level. At the multivariable level, the multiple linear regression (MLR) model for blood haemoglobin on a continuous scale was compared to the binary logistic regression model for blood haemoglobin, modeled as a binary outcome to identify the most suitable model based on the minimum deviance and Bayesian Information Criterion (BIC) criterion to examine factors associated with blood haemoglobin among pregnant women in Uganda at a 0.05 level of significance. The study established that the MLR model of the blood haemoglobin offered a better precision of identification of risk factors associated with it than the binary logistic regression model. The MLR analysis revealed that older pregnant women, having had more than one pregnancy (multigravida) and having at least three meals in a day, had a significant positive relationship with blood haemoglobin level, whereas being overweight had a significant negative relationship with blood haemoglobin level. The findings proved that attention to distributional analysis and the use of the MLR regression model rather than the binary logistic regression model improves the precision of identification of risk factors associated with blood haemoglobin among pregnant women. Body mass index (BMI) is a modifiable risk factor for blood haemoglobin concentration that can be addressed through lifestyle interventions. We recommend the use of the MLR model for modelling the risk factors for blood haemoglobin