Research Articles (Statistics & Population Studies)

Permanent URI for this collectionhttps://hdl.handle.net/10566/732

Browse

Now showing 1 - 14 of 14

A review and comparison of methods of parameter estimation and inference for heteroskedastic linear regression models
(Taylor and Francis Ltd., 2025) Farrar, Thomas J.; Blignaut, Rénette Julia; Luus, Retha; Steel, Sarel J.
This article reviews methods of parameter estimation and inference in the linear regression model under heteroskedasticity. Several approaches to feasible weighted least squares estimation of the parameter vector are reviewed, along with various heteroskedasticity-consistent covariance matrix estimators, which are usually designed with inference as the end goal. A Monte Carlo experiment is designed to evaluate the ability of the reviewed methods to estimate three quantities: the variances of the random errors, the parameter vector, and the standard error of the ordinary least squares estimator thereof. Results of the experiment show that the homoskedastic variance estimator performs well at estimating error variances even in the heteroskedastic data-generating processes studied. Feasible weighted least squares approaches perform best for estimation of the parameter vector, whereas heteroskedasticity-consistent covariance matrix estimators perform best for estimation of the standard error thereof. This motivates a search for a method that would perform well in all three respects.
Amplitude and phase changes in TESS photometry of five short period weak-lined T Tauri stars
(Oxford University Press, 2024) Koen, Chris
The statistical technique ‘complex demodulation’ (CDM) is used to track the amplitude and phase changes of periodicities in five naked T Tauri stars. The periodicities are most likely caused by dark spots on the stellar surfaces, which are rotated into and out of view. Two of the stars (CD-56 1438, CD-72 248) show two independent periodicities, probably due to being binary weak-lined T Tauri stars. Two different low-pass filters, operating, respectively, in the frequency and time domains, are used as part of the CDM methodology. Statistical aspects of the estimated amplitudes and phases are investigated in some detail: in particular, expressions are derived for standard errors and for possible biases. A large variety of different types of amplitude and phase changes are found, including approximately linear or quadratic, abrupt level shifts, pulses, and oscillatory. Long term changes in amplitudes are aperiodic, but mimic long-term cycles.
A mixed-methods study exploring women’s perceptions of terminology surrounding fertility and menstrual regulation in Côte d’Ivoire and Nigeria
(BMC, 2021) Sheehy, Grace; Omoluabi, Elizabeth; OlaOlorun, Funmilola M.
Women use various terms when discussing the management of their fertility and menstrual irregularities and may interpret the experience of ending a possible pregnancy in nuanced ways, especially when their pregnancy status is ambiguous. Our study aims to understand the terminology used to refer to abortion-like experiences (specifcally menstrual regulation and pregnancy removal), and the specifc scenarios that these practices encompass among women who reported doing something to bring back a late period or ending a pregnancy in Nigeria and Côte d’Ivoire.Our analysis draws upon surveys with women in Nigeria (n=1114) and Cote d’Ivoire (n=352). We also draw upon qualitative in-depth interviews with a subset of survey respondents in Anambra and Kaduna States in Nigeria, and Abidjan, Cote d’Ivoire (n=30 in both countries). We examine survey and interview questions that explored women’s knowledge of terminology pertaining to ending a pregnancy or bringing back a late period. Survey data were analyzed descriptively and weighted, and interview data were analyzed using inductive thematic analysis.
Using a mHealth system to recall and refer existing clients and refer community members with health concerns to primary healthcare facilities in South Africa: a feasibility study
(Taylor & Francis, 2020) Jordaan, Esme; Odendaal, Willem A.; Lewin, Simon A.
Background: Lay health workers (LHWs) are critical in linking communities and primary healthcare (PHC) facilities. Effective communication between facilities and LHWs is key to this role. We implemented a mobile health (mHealth) system to improve communication and continuity of care for chronically ill clients. The system focused on requests from facility staff to LHWs to follow up clients and LHW referrals of people who needed care at a facility. We implemented the system in two rural and semi-rural sub-districts in South Africa. Objective: To assess the feasibility of the mHealth system in improving continuity of care for clients in PHC in South Africa. Method: We implemented the intervention in 15 PHC facilities. The clerks issued recalls to LHWs using a tablet computer. LHWs used smartphones to receive these requests, communicate with clerks and refer people to a facility. We undertook a mixed-methods evaluation to assess the feasibility of the mHealth system. We analysed recall and referral data using descriptive statistics. We used thematic content analysis to analyse qualitative data from semi-structured interviews with facility staff and a researcher fieldwork journal. Results: Across the sub-districts, 2,204 clients were recalled and 628 (28%) of these recalls were successful. LHWs made 1,085 referrals of which 485 (45%) were successful.
Prerace medical screening and education reduce medical encounters in distance road races: SAFER VIII study in 153 208 race starters
(BMJ Publishing Group, 2018) Schwellnus, Martin; Swanevelder, Sonja; Derman, Wayne; Borjesson, Mats; Schwabe, Karen; Jordaan, Esme
OBJECTIVES To examine the efficacy and feasibility of an online prerace medical screening and educational intervention programme for reducing medical complications in long-distance races. METHODS This was an 8-year observational study of medical encounter rates among 153 208 Two Oceans race starters (21.1 and 56 km) in South Africa. After the first 4-year control (CON) period, we introduced an online prerace medical screening (based on European pre-exercise screening guidelines) and an automated educational intervention programme. We compared the incidence of medical encounters (per 1000 starters; all and serious life threatening) in the CON versus the 4-year intervention (INT) period. RESULTS In comparison to the CON period (2008–2011: 65 865 starters), the INT period (2012–2015: 87 343 starters) had a significantly lower incidence (adjusted for age group, sex, race distance) of all medical encounters by 29% (CON=8.6 (7.9–9.4); INT=6.1 (5.6–6.7), p<0.0001), in the 21.1 km race by 19% (CON=5.1 (4.4–5.9); INT=4.1 (3.6–4.8), p=0.0356) and in the 56 km race by 39% (CON=14.6 (13.1–16.3); INT=9.0 (7.9–10.1), p<0.0001). Serious life-threatening encounters were significantly reduced by 64% (CON=0.6 (0.5–0.9); INT=0.2 (0.1–0.4); p=0.0003) (adjusted for age group and sex). Registration numbers increased in the INT period (CON=81 345; INT=106 743) and overall % race starters were similar in the CON versus INT period. Wet-bulb globe temperature was similar in the CON and INT periods. CONCLUSION All medical encounters and serious life-threatening encounters were significantly lower after the introduction of a prescreening and educational intervention programme, and the programme was feasible.
Match injury incidence during the Super Rugby tournament is high: a prospective cohort study over five seasons involving 93 641 player-hours
(BMJ Publishing Group, 2018) Schwellnus, Martin; Jordaan, Esme; Janse van Rensburg, Charl; Bayne, Helen
OBJECTIVES: To determine the incidence and nature of injuries in the Super Rugby tournament over a 5-year period. METHODS: 482 male professional rugby union players from six South African teams participating in the Super Rugby tournament were studied (1020 player-seasons). Medical staff of participating teams (2012–2016 tournaments) recorded all time loss injuries (total injuries and match injuries) and exposure hours (93 641 total playing hours; 8032 match hours). Injury incidence, injured player proportion, severity (time lost), anatomical location, tissue type and activity/phase during which injury occurred are reported. RESULTS: The overall incidence of match injuries (per 1000 player-hours; 95% CI) for each year was as follows: 2012 (83.3; 69.4–99.2); 2013 (115.1; 98.7–133.5); 2014 (95.9; 80.8–113.1), 2015 (112.3; 96.6–129.9) and 2016 (93.2; 79.9–107.9). The injured player proportion for each year was as follows: 2012 (54.6%); 2013 (49.4%); 2014 (52.0%); 2015 (50.0%); and 2016 (39.8%). The thigh, knee, head/face and shoulder/ clavicle are the most frequently injured locations, and muscle/tendon and joint/ligament injuries account for the majority of injuries. Most injuries (79%) occur in contact situations, in particular during a tackle (54%). CONCLUSION: The incidence of match injuries and the injured player proportion in South African teams competing in the Super Rugby tournament is high. Match injury incidence is consistently higher than previously reported for senior male rugby players at elite/ professional level. Targeted risk management strategies are therefore needed in the Super Rugby tournament to manage risk of injury.
A search for p-mode pulsations in white dwarf stars using the Berkeley Visible Imaging Tube detector
(Oxford University Press, 2014) Kilkenny, David; Welsh, B.Y.; Koen, Chris; Gulbis, A.A.S.; Kotze, M.M.
We present high-speed photometry (resolution 0.1 s) obtained during the commissioning of the Berkely Visible Imaging Tube system on the Southern African Large Telescope (SALT). The observations were an attempt to search for very rapid p-mode oscillations in white dwarf stars and included three DA stars known to be g-mode pulsators (ZZ Cet, HK Cet and AF Pic), one other DA star (WD 1056-384) not known to be variable and one AM Cvn star (HP Lib). No evidence was found for any variations greater than about 1 mmag in amplitude (∼0.1 per cent) at frequencies in excess of 60 mHz (periods <17 s) in any of the target stars, though several previously known g-mode frequencies were recovered.
A detection threshold in the amplitude spectra calculated from Kepler data obtained during K2 mission.
(Oxford University Press, 2015) Baran, A.C.; Koen, Chris; Pokrzywka, B.
We present our analysis of simulated data in order to derive a detection threshold which can be used in the pre-whitening process of amplitude spectra. In case of ground-based data of pulsating stars, this threshold is conventionally taken to be four times the mean noise level in an amplitude spectrum. This threshold is questionable when space-based data are analysed. Our effort is aimed at revising this threshold in the case of continuous 90-d Kepler K2 phase observations. Our result clearly shows that a 95 per cent confidence level, common for ground observations, can be reached at 5.4 times the mean noise level and is coverage dependent. In addition, this threshold varies between 4.8 and 5.7, if the number of cadences is changed. This conclusion should secure further pre-whitening and helps to avoid over-interpretation of spectra of pulsating stars observed with the Kepler spacecraft during K2 phase. We compare our results with the standard approach widely used in the literature.
The analysis of irregularly observed stochastic astronomical time-series – I. Basics of linear stochastic differential equations
(Oxford University Press, 2005) Koen, Chris
The theory of low-order linear stochastic differential equations is reviewed. Solutions to these equations give the continuous time analogues of discrete time autoregressive time-series. Explicit forms for the power spectra and covariance functions of first- and second-order forms are given. A conceptually simple method is described for fitting continuous time autoregressive models to data. Formulae giving the standard errors of the parameter estimates are derived. Simulated data are used to verify the performance of the methods. Irregularly spaced observations of the two hydrogen-deficient stars FQ Aqr and NO Ser are analysed. In the case of FQ Aqr the best-fitting model is of second order, and describes a quasi-periodicity of about 20 d with an e-folding time of 3.7 d. The NO Ser data are best fitted by a first-order model with an e-folding time of 7.2 d.
Correlates of antenatal and postnatal care among tribal women in India
(Kamla-Raj Enterprises, 2012) Appunni, Sathiya Susuman
It is well established that those women who undergo full antenatal care check-up and give birth in a medical institution or whose delivery is attended by trained paramedical persons promotes child survival and reduces maternal mortality. In the light of the above, an attempt is made in this paper to relate the socio-economic and demographic characteristics of the currently married Schedule Tribes’ women in eight districts of Chhatisgarh with factors associated with antenatal and postnatal care. Data for this study were taken from District Level Household Survey on Reproductive and Child Health (DLHSRCH 2002), a representative sample of 1569 Scheduled Tribes’ currently married women aged 15-44, residing in eight districts of Chhatisgarh. Adjusted effects (odds ratios) analysis has been used to find out the effects of antenatal and post-natal care on institutional delivery in Chhatisgarh. It is observed that majority of the Scheduled Tribes’ women, about 84 percent, have a low standard of living. Also, 74 per cent of the Scheduled Tribes’ women are illiterate. The finding of the adjusted effects (odds ratio) shows that giving birth in the medical institution for the Scheduled Tribes’ women who received full antenatal check up is 2.5 times higher than those women who did not receive any antenatal check-up. It suggests that majority of the currently married Scheduled Tribe women have low standard of living there is a need to improve their economic standard so that they can fulfill their basic needs.
Multiple regression of GRB luminosity on light-curve properties
(Oxford University Press, 2009) Koen, Chris
Statistical relations between the luminosity L of gamma-ray burster (GRBs) and several of their other observable properties have been discovered. Four of these properties are considered here: the spectral lag; a measure of the light-curve variability; peak energy and the minimum light-curve rise time. Data are taken from a tabulation by Schaefer. Log–log regression of L on various combinations of the four light-curve properties, as well as redshift, is considered, using conventional multiple linear regression, and multiple errors-in-variables regression. Several cases are found in which the regression coefficients of more than one luminosity indicators are significant. In particular, the simultaneous regression of luminosity on peak energy, spectral lag, minimum rise time and redshift is meaningful.
Evidence for rapid evolution of periodic variations in an ultracool dwarf
(Oxford University Press, 2006) Koen, Chris
The results of three short photometric monitoring runs on the L0 dwarf 2MASS J06050196−2342270, during three consecutive nights, are presented. The observations show the persistent presence of a 2.4-h period, with an IC band amplitude which decreased from 27 to 11 mmag over the three nights. The amplitude in the RC band appeared to be substantially smaller than in the IC band.
On multiple classes of gamma-ray bursts, as deduced from autocorrelation functions or bivariate duration/hardness ratio distributions
(Oxford University Press, 2012) Koen, Chris; Bere, A.
Autocorrelation functions (ACFs) of 119 gamma-ray bursts (GRBs) monitored by the Burst Alert Telescope (BAT) on Swift are calculated. Contrasting with previous results for smaller numbers of bursts from other missions, the widths of the ACFs are not bimodally distributed. Although the distribution appears slightly asymmetrical, underlying mixtures of distributions can also probably be ruled out. Factors contributing to differences between the results presented here, and those in the literature, may include the differences in energy passbands used, and the superior sensitivity of the BAT instrument (which affects e.g. the redshift distribution of the detected GRB). The second part of the paper is concerned with the fitting of mixtures of bivariate Gaussians to the joint duration/hardness ratio data of 325 GRBs. A careful analysis confirms that a three-component mixture is the statistically most acceptable, but it is shown that the implied marginal distribution of the hardness ratios does not fit the data very well. It is also stressed that mixture components cannot automatically be assumed to represent different classes of GRBs. The point is illustrated by showing two substantially different, but statistically almost equivalent, mixture models for the distribution of 571 BAT-determined GRB durations.
Statistical tests for changes in the amplitude, frequency or phase of a sinusoidal variation
(Oxford University Press, 2009) Koen, Chris
The problem considered is that of testing for small changes over time in the properties of a sinusoidal signal contaminated by noise. Two test statistics are proposed. The first is a frequency domain statistic based on the spectrum of the data, while the second is a time domain statistic due to Nyblom. Power studies are used to show that the Nyblom statistic is generally the better. The form of the Nyblom statistic appropriate for phase-change tests also performs well in testing for frequency changes.

Browse

Recent Submissions