Browsing by Subject "Regression analysis"

Now showing 1 - 6 of 6
  • Thumbnail Image
    Item
    Open Access
    Antiretroviral treatment cohort analysis using time-updated CD4 counts: assessment of bias with different analytic methods
    (Public Library of Science, 2011) Kranzer, Katharina; Lewis, James J; White, Richard G; Glynn, Judith R; Lawn, Stephen D; Middelkoop, Keren; Bekker, Linda-Gail; Wood, Robin
    BACKGROUND: Survival analysis using time-updated CD4+ counts during antiretroviral therapy is frequently employed to determine risk of clinical events. The time-point when the CD4+ count is assumed to change potentially biases effect estimates but methods used to estimate this are infrequently reported. METHODS: This study examined the effect of three different estimation methods: assuming i) a constant CD4+ count from date of measurement until the date of next measurement, ii) a constant CD4+ count from the midpoint of the preceding interval until the midpoint of the subsequent interval and iii) a linear interpolation between consecutive CD4+ measurements to provide additional midpoint measurements. Person-time, tuberculosis rates and hazard ratios by CD4+ stratum were compared using all available CD4+ counts (measurement frequency 1-3 months) and 6 monthly measurements from a clinical cohort. Simulated data were used to compare the extent of bias introduced by these methods. RESULTS: The midpoint method gave the closest fit to person-time spent with low CD4+ counts and for hazard ratios for outcomes both in the clinical dataset and the simulated data. CONCLUSION: The midpoint method presents a simple option to reduce bias in time-updated CD4+ analysis, particularly at low CD4 cell counts and rapidly increasing counts after ART initiation.
  • Thumbnail Image
    Item
    Open Access
    BDNF polymorphisms are linked to poorer working memory performance, reduced cerebellar and hippocampal volumes and differences in prefrontal cortex in a Swedish elderly population
    (Public Library of Science, 2014) Brooks, Samantha J; Nilsson, Emil K; Jacobsson, Josefin A; Stein, Dan J; Fredriksson, Robert; Lind, Lars; Schiöth, Helgi B
    BACKGROUND: Brain-derived neurotrophic factor (BDNF) links learning, memory and cognitive decline in elderly, but evidence linking BDNF allele variation, cognition and brain structural differences is lacking. METHODS: 367 elderly Swedish men (n = 181) and women (n = 186) from Prospective Investigation of the Vasculature in Uppsala seniors (PIVUS) were genotyped and the BDNF functional rs6265 SNP was further examined in subjects who completed the Trail Making Task (TMT), verbal fluency task, and had a magnetic resonance imaging (MRI) scan. Voxel-based morphometry (VBM) examined brain structure, cognition and links with BDNF. RESULTS: The functional BDNF SNP (rs6265,) predicted better working memory performance on the TMT with positive association of the Met rs6265, and was linked with greater cerebellar, precuneus, left superior frontal gyrus and bilateral hippocampal volume, and reduced brainstem and bilateral posterior cingulate volumes. CONCLUSIONS: The functional BDNF polymorphism influences brain volume in regions associated with memory and regulation of sensorimotor control, with the Met rs6265 allele potentially being more beneficial to these functions in the elderly.
  • Thumbnail Image
    Item
    Open Access
    A detailed investigation of the linear model and some of its underlying assumptions
    (1977) Coutsourides, Dimitris; Troskie, Casper G
    The purpose of this thesis is to provide a study of the linear model. The whole work has been split into 6 chapters. In Chapter 1 we define and examine the two linear models, i.e. the regression and the correlation model. More specifically we show that the regression model is the conditional version of the correlation model. In Chapter 2 we deal with the problem of multicollinearity. We investigate the sources of near singularities, we give some methods of detecting the multicollinearity, and we state briefly methods for overcoming this problem. In Chapter 3 we consider the least squares method with restrictions, and we dispose of some tests for testing the linear restrictions. The theory concerning the sign of least squares estimates is discussed, then we deal with the method for augmenting existing data. Chapter 4 is mainly devoted to ridge regression. We state methods for selecting the best estimate for k. Some extensions are given dealing with the shrinkage estimators and the linear transforms of the least squares. In Chapter 5 we deal with the principal components, and we give methods for selecting the best subset of principal components. Much attention was given to a method called fractional rank and latent root regression analysis. In Chapter 6 comparisons were performed between estimators previously mentioned. Finally the conclusions are stated.
  • Thumbnail Image
    Item
    Open Access
    The estimation of missing values in hydrological records using the EM algorithm and regression methods
    (1988) Makhuvha, Tondani; Zucchini, Walter; Sparks, Ross S
    The objective of this thesis is to review existing methods for estimating missing values in rainfall records and to propose a number of new procedures. Two classes of methods are considered. The first is based on the theory of variable selection in regression. Here the emphasis is on finding efficient methods to identify the set of control stations which are likely to yield the best regression estimates of the missing values in the target station. The second class of methods is based on the EM algorithm, proposed by Dempster, Laird and Rubin (1977). The emphasis here is to estimate the missing values directly without first making a detailed selection of control stations. All "relevant" stations are included. This method has not previously been applied in the context of estimating missing rainfall values.
  • Thumbnail Image
    Item
    Open Access
    Quality assurance in high schools through regression analysis
    (1997) Watermeyer, John Westerton; Gilmour, James David; Dunne, Tim T
    School Effectiveness is a relatively new and poorly defined domain for which a structure is proposed, to facilitate future discussion. Three fields within School Effectiveness are identified, namely School Effectiveness Research (SER), School Improvement (SI) and Quality Assurance (QA). Three divisions are identified within each field on the basis of various criteria. SER has methodological generations, Sl is classified by decade, and three themes of QA are described, including performance indicators (Pis). A definition of effectiveness in terms of regression lines is described and the concept of added value or adjusted achievement developed. This study is concerned with the development of Pis for use within a single school to monitor and promote improvement. The context of the study (a model C senior high school in a predominantly white southern suburb of Cape Town) and the data collected is described before a review is made of some of the analyses which could be used to monitor effectiveness. A technique whereby pupil achievement is adjusted (for prior achievement and other background variables) and the residuals (or adjusted achievement) derived from the regressions investigated with one-way A NOVAs is described and tested using various models and subjects. With respect to groups, it is proposed that statistical significance of differences between mean residuals could be used as a PI. With respect to individual pupils, educators could set their own criterion for investigating cases where adjusted achievement is very large or very small. Statistical significance requires interpretation, however, and the role of professional judgement in modelling and monitoring adjusted achievement is discussed. The view that techniques such as regression analysis can only indicate when professional investigation and intervention might be necessary is stressed. It would seem unwise to rank teachers or subjects on the basis of adjusted achievement.
  • Thumbnail Image
    Item
    Open Access
    Treatment Interruption and Variation in Tablet Taking Behaviour Result in Viral Failure: A Case-Control Study from Cape Town, South Africa
    (Public Library of Science, 2011) Ncaca, Lisa-Noelle; Kranzer, Katharina; Orrell, Catherine
    BACKGROUND: Understanding of the impact of non-structured treatment interruption (TI) and variation in tablet-taking on failure of first-line antiretroviral therapy (ART) is limited in a resource-poor setting. METHODS: A retrospective matched case-control analysis. Individuals failing ART were matched by time on ART with 4 controls. Viral load (VL) and CD4 count were completed 4-monthly. Adherence percentages, from tablet returns, were calculated 4-monthly (interval) and from ART start (cumulative). Variation between intervals and TI (>27 days off ART) were recorded. Conditional multivariate logistic regression analysis was performed to estimate the effect of cumulative adherence <90%, at least one episode of adherence variation >10% and TI on virological failure. Age, gender, baseline log VL and CD4 were included as possible confounders in the multivariate model. RESULTS: 244 patients (44 cases, 200 controls) were included. Median age was 32 years (IQR28-37), baseline CD4 108 cells/mm3 (IQR56-151), VL 4.82 log (IQR4.48-5.23). 94% (96% controls, 86% failures) had cumulative adherence >90%. The odds of failure increased 3 times (aOR 3.01, 95%CI 0.81-11.21) in individuals with cumulative adherence <90%, 2.2 times (aOR 2.20, 95%CI 1.04-4.64) in individuals with at least one episode of fluctuating adherence of >10% and 4.01 times (aOR 4.01, 95%CI 1.45-11.10) in individuals with TIs. For individuals with TI and cumulative adherence >95%, the odds of failing were 5.65 (CI 1.40-22.85). CONCLUSION: It is well known that poor cumulative adherence increases risk of virological failure, but less well understood that TI and variations in tablet-taking also play a key role, despite otherwise excellent adherence.