Logit Link Function Research Articles

Dead before detection: addressing the effects of left truncation on survival estimation and ecological inference for neonates

SummaryNeonate survival is a key life history trait, yet remains challenging to measure in wild populations because neonates can be difficult to capture at birth. Estimates of survival from neonates that are opportunistically captured might be inaccurate because some individuals die before sampling, resulting in data that are left truncated. The resulting overestimation of survival rates can further affect ecological inference through biased estimates of covariate effects in survival models, yet is not addressed in most studies of animal survival. Here, we quantify the effects of left truncation on survival estimates and subsequent ecological inference.Vaginal implant transmitters (VITs) enable capture of ungulates at birth, yielding data without left truncation. The effects of left truncation on survival estimation were quantified using age‐dependent survival models for VIT and opportunistically captured neonatal deer. Differences in daily survival rates (DSRs) and cumulative survival probability were calculated for the first 70 days of life. In addition, left truncation was simulated by removing fawns that died during the first 1 or 2 days of life from the VIT‐caught sample, isolating the effect of left truncation.Cumulative probability of survival during the first 70 days of life was overestimated by 7–23% for fawns caught opportunistically compared with those caught by VIT, depending on model design. Differences in DSRs were large at age 1 day, but had converged by age 30 days. Simulated left truncation resulted in overestimates of survival of up to 31%. Model selection and covariate coefficients were strongly affected by left truncation, producing spurious ecological inference, including changes to sign and/or magnitude of inferred effects of all covariates.We recommend (i) every effort be made to capture neonates; (ii) consistent capture methods, using at least in part non‐truncating techniques, be implemented across years and study areas; and (iii) exclusion of left‐truncated data from survival estimates until DSRs converge with those calculated from non‐truncated data. This work emphasizes the importance of accounting for left truncation in survival estimation for any species with strong age‐dependent survival in order to prevent biased conclusions produced by sampling method rather than true ecological effects.

The impact of methadone maintenance therapy on hepatitis C incidence among illicit drug users.

To determine the relationship between methadone maintenance therapy (MMT) and hepatitis C (HCV) seroconversion among illicit drug users. A generalized estimating equation model assuming a binomial distribution and a logit-link function was used to examine for a possible protective effect of MMT use on HCV incidence. Data from three prospective cohort studies of illicit drug users in Vancouver, Canada between 1996 and 2012. A total of 1004 HCV antibody-negative illicit drug users stratified by exposure to MMT. Baseline and semi-annual HCV antibody testing and standardized interviewer-administered questionnaire soliciting self-reported data relating to drug use patterns, risk behaviors, detailed socio-demographic data and status of active participation in an MMT program. One hundred and eighty-four HCV seroconversions were observed for an HCV incidence density of 6.32 [95% confidence interval (CI) = 5.44-7.31] per 100 person-years. After adjusting for potential confounders, MMT exposure was protective against HCV seroconversion [adjusted odds ratio (AOR) = 0.47; 95% CI = 0.29-0.76]. In subanalyses, a dose-response protective effect of increasing MMT exposure on HCV incidence (AOR = 0.87; 95% CI = 0.78-0.97) per increasing 6-month period exposed to MMT was observed. Participation in methadone maintenance treatment appears to be highly protective against hepatitis C incidence among illicit drug users. There appears to be a dose-response protective effect of increasing methadone exposure on hepatitis C incidence.

Pre-transplant immune state defined by serum markers and alloreactivity predicts acute rejection after living donor kidney transplantation.

Acute rejection (AR) remains a major cause for long-term kidney allograft failure. Reliable immunological parameters suitable to define the pre-transplant immune state and hence the individual risk of graft rejection are highly desired to preferably adapt the immunosuppressive regimen in advance. Donor and third party alloreactivities were determined by mixed lymphocyte cultures. Soluble forms of CD25, CD30, and CD44 were detected in patients' serum by ELISA. Various lymphocyte subpopulations were measured using flow cytometry. All patients received triple immunosuppression (tacrolimus/mycophenolate mofetil/steroids) and were grouped according to biopsy results within the first year: rejection-free (RF, n=13), borderline (BL, n=5), or acute rejection (AR, n=7). Patients with AR showed the highest pre-transplant alloreactivities and serum levels (sCD25/sCD30/sCD44) according to the pattern RF<BL<AR. Relying on serum analysis only, multivariate logistic regression (logit link function) yielded a prognostic score for prediction of rejection with 75.0% sensitivity and 69.2% specificity. Patients with rejection showed markedly higher pre-transplant frequencies of CD4(+) /CD8(+) T cells lacking CD28, but lower numbers of CD8(+) CD161(bright) T cells and NK cells than RF individuals. Pre-transplant immune state defined by alloreactivity, serum markers, and particular lymphocyte subsets seems to correlate with occurrence of graft rejection after kidney transplantation. A prognostic score based on pre-transplant serum levels has shown great potential for prediction of rejection episodes and should be further evaluated.

A PARAMETER‐DRIVEN LOGIT REGRESSION MODEL FOR BINARY TIME SERIES

We consider a parameter‐driven regression model for binary time series, where serial dependence is introduced by an autocorrelated latent process incorporated into the logit link function. Unlike in the case of parameter‐driven Poisson log‐linear or negative binomial logit regression model studied in the literature for time series of counts, generalized linear model (GLM) estimation of the regression coefficient vector, which suppresses the latent process and maximizes the corresponding pseudo‐likelihood, cannot produce a consistent estimator. As a remedial measure, in this article, we propose a modified GLM estimation procedure and show that the resulting estimator is consistent and asymptotically normal. Moreover, we develop two procedures for estimating the asymptotic covariance matrix of the estimator and establish their consistency property. Simulation studies are conducted to evaluate the finite‐sample performance of the proposed procedures. An empirical example is also presented.

MODELING RATIOS WITH POTENTIAL ZERO-INFLATION TO ASSESS SOIL NEMATODE COMMUNITY STRUCTURE

The southern root-knot nematode (SRKN) and the weedy perennials, yellow nutsedge (YNS) and purple nutsedge (PNS) are simultaneously-occurring pests in the irrigated agricultural soils of southern New Mexico. Previous research has characterized SRKN, YNS and PNS as a mutually beneficial pest complex and has shown their enhanced population growth and survival when they occur together. In addition, it was shown that the density of nutsedge in a field could be used as a predictor of SRKN juveniles in the soil. In addition to SRKN, which is the most harmful of the plant parasitic nematodes, in southern New Mexico other species or categories of nematodes were identified and counted. Some of them are not as damaging to crop plants as SRKN, and some of them may be essential for soil health. The nematode species could be grouped into categories according to trophic level (what nematodes eat) and herbivore feeding behavior (how herbivore nematodes eat). Then three ratios of counts each were calculated for trophic and feeding behavior categories to investigate the soil nematode community structure. These proportions were modeled as functions of the weed hosts YNS and PNS by generalized linear regression models using the logit link function and three distributions: the Binomial, Zero-Inflated Binomial (ZIB) and Binomial Hurdle (BH). The latter two were used to account for potential high proportions of zeroes in the data. Formulas for the probability mass functions and moments were developed for the ZIB and BH. The SAS NLMIXED procedure was used to fit models for each of three sampling dates (May, July and September) in the two years of an alfalfa field study. General results showed that the Binomial generally provided the best fit, indicating lower zero-inflation than expected, but that ZIB and BH are often comparable. Importance of YNS and PNS predictors varied over sample dates and ratios. Specific results for one selected ratio illustrate the differences in estimated probabilities between Binomial, ZIB and BH distributions as YNS counts increase.

D-optimal designs for beta regression models with single predictor

The problem of construction of D-optimal designs for beta regression models involving one predictor is considered for the mean-precision parameterization suggested by Ferrari and Cribari-Neto [Beta regression for modelling rates and proportions. J Appl Stat. 2004;31:799–815]. Here we use the logit link function for the mean sub-model. These designs are presented and discussed for unrestricted as well as restricted design regions by considering the precision parameter as (1) a known constant and (2) an unknown constant. Efficiency comparison of obtained designs with commonly used equi-weighted, equi-spaced designs is made to recommend designs for practical use. Real-life applications are given to show the usefulness of these designs.

Evaluation of gonadotropin-releasing hormone hydrogen chloride at 3 doses with prostaglandin F2α for fixed-time artificial insemination in dairy cows

The objectives of the current study were to evaluate the efficacy and field safety of GnRH HCl administered at 3 doses in fixed-time artificial insemination (FTAI) programs (Ovsynch) in dairy cows. A common protocol was conducted at 6 commercial dairies. Between 188 and 195 cows were enrolled at each site (total enrolled = 1,142). Cows had body condition scores ≥2 and ≤4, were between 32 to 140d in milk, and were clinically healthy. Within pen and enrollment day (enrollment cohort), cows were assigned randomly in blocks of 4 to each of 4 treatments: (1) 25mg of PGF2α on d 7 with FTAI 72±2h later (control); (2) 100μg of GnRH on d 0, d 7 a dose of 25mg of PGF2α, and the second administration of 100μg of GnRH (T100) administered either at 48±2h (d 9) after PGF2α with FTAI 24±2h later or 56±2h (d 9) after PGF2α and FTAI 17±2h later; (3) same as T100 with both injections of 150μg of GnRH (T150); and (4) same as T100 with both injections of 200μg of GnRH (T200). Three sites selected the first option and 3 sites selected the second option for the timing of the second injection of all doses of GnRH. Cows were observed daily for signs of estrus and adverse clinical signs. Cows not returning to estrus had pregnancy diagnosis between 42 and 65d following FTAI. Pregnancies per FTAI (P/FTAI) were analyzed as a binary variable (1 = pregnant, 0 = not pregnant) using a generalized linear mixed model with a binomial error distribution and a logit link function. The statistical model included fixed effects for treatment, random effects of site, site by treatment, enrollment cohort within site, and residual. Parity (first vs. second or greater) was included as a covariate. For demonstration of effectiveness, α=0.05 and a 2-tailed test were used. Fifty-two cows were removed from the study because of either deviation from the protocol, injury, illness, culling, or death. Among the remaining 1,090 cows, 33.9% were primiparous and 66.1% were multiparous. Back-transformed least squares means for P/FTAI were 17.1, 27.3, 29.1, and 32.2% for control, T100, T150 and T200, respectively. The P/FTAI for each GnRH dose differed from that of the control. No differences were detected in P/FTAI between GnRH doses. No treatment-related adverse events were observed. Mastitis was the most frequently observed adverse clinical sign, followed by lameness and pneumonia. This study documents the efficacy and safety of doses of 100 to 200μg of GnRH as the HCl salt when used in Ovsynch programs.

Effect of carryover and presampling procedures on the results of real-time PCR used for diagnosis of bovine intramammary infections with Streptococcus agalactiae at routine milk recordings

The use of PCR tests as diagnostics for intramammary infections (IMI) based on composite milk samples collected in a non-sterile manner at milk recordings is increasing. Carryover of sample material between cows and non-aseptic PCR sampling may be incriminated for misclassification of IMI with Streptococcus agalactiae (S. agalactiae) in dairy herds with conventional milking parlours. Misclassification may result in unnecessary costs for treatment and culling. The objectives of this study were to (1) determine the effect of carryover on PCR-positivity for S. agalactiae at different PCR cycle threshold (Ct) cut-offs by estimating the between-cow correlation while accounting for the milking order, and (2) evaluate the effect of aseptic presampling procedures (PSP) on PCR-positivity at the different Ct-value cut-offs.The study was conducted in four herds with conventional milking parlours at routine milk recordings. Following the farmers’ routine pre-milking preparation, 411 of 794 cows were randomly selected for the PSP treatment. These procedures included removing the first streams of milk and 70% alcohol teat disinfection. Composite milk samples were then collected from all cows and tested using PCR. Data on milking order were used to estimate the correlation between consecutively milked cows in each milking unit. Factors associated with the PCR-positivity for S. agalactiae were analyzed using generalized estimating equations assuming a binomially-distributed outcome with a logit link function. Presampling procedures were only significant using cut-off 37. A first-order autoregressive correlation structure provided the best correlation between consecutively milked cows. The correlation was 13%, 11%, 9% at cut-offs <40, 37, and 34, respectively. PSP did not reduce the odds of cows being PCR-positive for S. agalactiae.In conclusion, carryover and non-aseptic sampling affected the PCR results and should therefore be considered when samples from routine milk recordings are used. In relative terms, higher cut-offs resulted in higher between-cow correlation, but the absolute amount of carryover may not be affected although this was not tested.

Correlates of sexual activity versus non-activity of incoming first-year students at a South African university

In order to contribute to the design of more effective programmes to curb the spread of HIV at tertiary institutions, this study compares the profile of students who are sexually active versus those who are not yet sexually active when entering university. The study was conducted among three cohorts of first-year university students at the University of the Western Cape from 2007 to 2009. A range of correlates of sexual activity versus non-activity were explored using logistic regression analysis. The predicted probabilities of the logit link function depict marked differences between genders and racial groups. Males were more likely than females to be sexually active when entering university and the Black racial group was more likely than other racial groups to be sexually active. Risk-taking behaviours such as smoking, alcohol and drug use were shown to increase the likelihood of being sexually active when entering university, as did indicators of depression and suicidal ideation. The results indicate that religion plays an important role in influencing sexual behaviour, highlighting the important potential role that religious organisations can play in addressing HIV risk at tertiary institutions.

Prioritizing School Intramural and Interscholastic Programs Based on Observed Physical Activity

To examine which school sports engage children in more physical activity. Observational, cross-sectional study examining differences between intramural (IM) and interscholastic (IS) sports. Athletic facilities at two schools with IM sports and two schools with IS sports in Wake County, North Carolina. Middle-school children (N = 6735). Percentage of children observed in sedentary, moderate, and vigorous activity assessed by the System for Observation Play and Leisure Among Youth (SOPLAY). Energy expenditure and physical activity intensity were also estimated by using MET values. T-tests; generalized linear model using cumulative logit link function. IM sports had higher MET values than IS sports (t = -3.69, p < .001), and IM sports ranked in four of the top five sports in terms of average MET values. Regression models found a significant interaction between school sport delivery model and gender, with boys significantly less physically active in IS programs than boys in IM programs (B = -.447, p < .001) but more physically active (B = .359, p <.001) than girls in IM sports. Regardless of sport type, IM sports generated more physical activity than IS sports among boys but not girls. Soccer, basketball, and track, regardless of school delivery approach, provide the highest physical activity levels. Results suggest that school administrators consider reassessing their programs to more efficiently use diminishing resources to increase students' physical activity levels.

Case Managers’ Perceptions of Consumer Work Readiness and Association With Pursuit of Employment

OBJECTIVE The purpose of this study was to determine whether case managers' perceptions of consumers' work readiness-in the areas of the individual consumer's hygiene, social skills, medication adherence, psychiatric symptoms, and substance use-influence pursuit of employment. METHODS A cross-sectional survey study design was used to assess the role of staff perceptions in consumers' pursuit of employment. A hierarchical generalized linear model with a logit link function was used to analyze data from 1,556 consumers nested within data for 113 case manager caseloads. RESULTS Case managers' perception of work readiness was significantly associated with pursuit of employment. CONCLUSIONS The clients of case managers with less stringent criteria for pursuing employment were more likely to take positive steps toward employment.

Localized Prostate Cancer Detection with18F FACBC PET/CT: Comparison with MR Imaging and Histopathologic Analysis

To characterize uptake of 1-amino-3-fluorine 18-fluorocyclobutane-1-carboxylic acid ((18)F FACBC) in patients with localized prostate cancer, benign prostatic hyperplasia (BPH), and normal prostate tissue and to evaluate its potential utility in delineation of intraprostatic cancers in histopathologically confirmed localized prostate cancer in comparison with magnetic resonance (MR) imaging. Institutional review board approval and written informed consent were obtained for this HIPAA-compliant prospective study. Twenty-one men underwent dynamic and static abdominopelvic (18)F FACBC combined positron emission tomography (PET) and computed tomography (CT) and multiparametric (MP) 3-T endorectal MR imaging before robotic-assisted prostatectomy. PET/CT and MR images were coregistered by using pelvic bones as fiducial markers; this was followed by manual adjustments. Whole-mount histopathologic specimens were sliced with an MR-based patient-specific mold. (18)F FACBC PET standardized uptake values (SUVs) were compared with those at MR imaging and histopathologic analysis for lesion- and sector-based (20 sectors per patient) analysis. Positive and negative predictive values for each modality were estimated by using generalized estimating equations with logit link function and working independence correlation structure. (18)F FACBC tumor uptake was rapid but reversible. It peaked 3.6 minutes after injection and reached a relative plateau at 15-20 minutes (SUVmax[15-20min]). Mean prostate tumor SUVmax(15-20min) was significantly higher than that of the normal prostate (4.5 ± 0.5 vs 2.7 ± 0.5) (P < .001); however, it was not significantly different from that of BPH (4.3 ± 0.6) (P = .27). Sector-based comparison with histopathologic analysis, including all tumors, revealed sensitivity and specificity of 67% and 66%, respectively, for (18)F FACBC PET/CT and 73% and 79%, respectively, for T2-weighted MR imaging. (18)F FACBC PET/CT and MP MR imaging were used to localize dominant tumors (sensitivity of 90% for both). Combined (18)F FACBC and MR imaging yielded positive predictive value of 82% for tumor localization, which was higher than that with either modality alone (P < .001). (18)F FACBC PET/CT shows higher uptake in intraprostatic tumor foci than in normal prostate tissue; however, (18)F FACBC uptake in tumors is similar to that in BPH nodules. Thus, it is not specific for prostate cancer. Nevertheless, combined (18)F FACBC PET/CT and T2-weighted MR imaging enable more accurate localization of prostate cancer lesions than either modality alone.

A general regression framework for a secondary outcome in case-control studies

Modern case-control studies typically involve the collection of data on a large number of outcomes, often at considerable logistical and monetary expense. These data are of potentially great value to subsequent researchers, who, although not necessarily concerned with the disease that defined the case series in the original study, may want to use the available information for a regression analysis involving a secondary outcome. Because cases and controls are selected with unequal probability, regression analysis involving a secondary outcome generally must acknowledge the sampling design. In this paper, the author presents a new framework for the analysis of secondary outcomes in case-control studies. The approach is based on a careful re-parameterization of the conditional model for the secondary outcome given the case-control outcome and regression covariates, in terms of (a) the population regression of interest of the secondary outcome given covariates and (b) the population regression of the case-control outcome on covariates. The error distribution for the secondary outcome given covariates and case-control status is otherwise unrestricted. For a continuous outcome, the approach sometimes reduces to extending model (a) by including a residual of (b) as a covariate. However, the framework is general in the sense that models (a) and (b) can take any functional form, and the methodology allows for an identity, log or logit link function for model (a).

Disparities in Use of Laparoscopic Hysterectomies: A Nationwide Analysis

Study ObjectiveTo determine patient and hospital characteristics that were associated with undergoing laparoscopic hysterectomy compared with abdominal hysterectomy. DesignCanadian Task Force Classification II-3. MethodsIn this retrospective cohort study, we analyzed the 2010 Healthcare Cost and Utilization Project Nationwide Inpatient Sample database. All women who underwent laparoscopic or abdominal hysterectomy for either menorrhagia or leiomyoma were included based on International Classification of Diseases, Ninth Revision coding. A linear model with binomial distribution and logit link function was used to determine patient and hospital characteristics associated with hysterectomy approach. Main ResultsA total of 32 436 patients were included in this study. Of these, 32% patients underwent laparoscopic hysterectomies, and 67% underwent abdominal hysterectomies. With regard to patient characteristics, women younger than 35 years old were more likely to undergo laparoscopic hysterectomy when compared with each of the other age categories (p < .001). White women were more likely to undergo laparoscopic hysterectomy than black women, Hispanic women, or women classified as “other” races (p < .001 for all comparisons). With regard to median income, patients from the lowest national quartile were less likely to undergo laparoscopic hysterectomy when compared with each of the other 3 national quartiles for income (p = .01, p < .001, p = .001, respectively). Payment by private insurance was associated with laparoscopic hysterectomy when compared with payment by Medicare or payment by insurance category “other” (p < .001 for both). With regard to hospital characteristics, hospitals in the Northeast were more likely to have laparoscopic hysterectomies than hospitals in the Midwest or South (p < .001 for both comparisons); urban hospitals were more likely than rural hospitals (p < .001); teaching hospitals were more likely than nonteaching hospitals (p < .001); and government-owned hospitals were less likely than private, nonprofit or private, investor owned (p < .001 for both comparisons). ConclusionsDespite the increased popularity of and training in laparoscopic hysterectomies, there remains an obvious disparity in its delivery with regard to patient and hospital characteristics. Further investigation is needed on the etiology of this disparity and interventions that may alleviate it.

Trajectory of Suicidal Ideation in Relation to Perceived Overweight From Adolescence to Young Adulthood in a Representative United States Sample

To examine the trajectory of suicidal ideation in relation to perceived overweight from adolescence to young adulthood. Four waves (1995-2008) of the National Longitudinal Study of Adolescent Health that prospectively followed up a nationally representative sample of United States middle and high school students (N= 6,504) were analyzed using hierarchical generalized linear models with a logit link function. Overall, suicidal ideation tended to decrease with age (logit coefficient [coef.]=-.07, p<.001). Participants who perceived themselves as overweight were more likely to think about committing suicide, especially in females, even after controlling for rigorously measured depressive symptoms (coef.= .14, p < .01). Girls were more likely than boys to have suicidal ideation at age 10 (coef.=-.31, p < .001). However, the association between suicidal ideation and perceived overweight became weaker and the prevalence of suicidal ideation became lower as participants grew older. The downward trend in suicidal ideation was steeper in females than in males (coef.= .03, p < .001), with males actually showing a higher prevalence of suicidal ideation than females around 20-30 years of age. Random effects of age, depression, and perceived overweight were insignificant, which indicated that participants did not vary significantly in terms of effects of these variables on suicidal ideation. The results of this study underline the need for development of family, school, and community interventions to address body weight perception to reduce suicidal ideation and attempts, especially among female teenagers.

Geographical Distribution of Dental Caries Prevalence and Associated Factors in Young Adults in Finland

Oral health of the young has been reported to be associated with the place of residence, due to differences in socio-economic status. The results of studies on the effect of fluoride in drinking water on caries prevalence have been contradictory. The main aim of our study was to investigate the geographical distribution of dental caries prevalence in Finland and analyse possible regional, associated factors. Oral health of 13,564 men born between 1990 and 1992 was screened using a method based on criteria of the World Health Organization for epidemiological studies by 15 calibrated dentists in 20/24 garrisons of the Finnish Defence Forces in 2011. Mean DMFT and DT values in provinces were calculated. Multilevel analysis was applied to the data using generalized linear mixed models and a logit link function. The binary outcome variable was the cariological treatment need (yes/no) and the garrison was treated as random effect. Mean DMFT and DT values varied significantly between provinces in Finland. Increased levels of fluoride in drinking water, Swedish as the main language in the municipality, and living in urban areas were protective factors against cariological treatment need. Dentist density did not affect caries prevalence. The geographical distribution of dental caries further supported estimates of the protective effects of high fluoride levels in drinking water, Swedish as the main language in the municipality, as well as living in urban areas. Effect of fluoride on caries prevalence still exists, and association with the socio-economic status was confirmed.

Why do people change routes? Impact of information services

PurposeThe aim of this study is to identify factors that influence drivers' route choice response to travel time information about both the expressway and local streets provided by variable message signs on arterial roads.Design/methodology/approachA stated preference questionnaire survey was conducted to collect behavioral data. The generalized estimating equations (GEEs) method with a logit link function was used to model driver response and account for correlations within repeated observations from the same respondent. Four GEEs‐based estimations with different working correlation structures were conducted and compared with each other as well as the conventional maximum likelihood estimation.FindingsDriving experiences, expressway delays, causes of delay, and the number of traffic lights on local streets are factors influencing route choice decisions. A new finding is that there exist differences in response behavior among employer‐provided car, taxi and private car drivers. On the modeling aspect, the exchangeable structure was the most appropriate in this study.Research limitations/implicationsThis study indicates the effectiveness and appropriateness of the GEEs method and suggests further examination of GEEs' performance.Practical implicationsThe route choice probability model established by this study will facilitate better investment, design and assessment of dynamic information services in transportation management.Originality/valueThe dynamic information this study concerns has rarely been addressed in the literature. Little literature to date has applied the GEEs method in information response modeling. This study reaches solider conclusions about the GEEs method.

Genomic prediction of dichotomous traits with Bayesian logistic models

Bayesian methods are a popular choice for genomic prediction of genotypic values. The methodology is well established for traits with approximately Gaussian phenotypic distribution. However, numerous important traits are of dichotomous nature and the phenotypic counts observed follow a Binomial distribution. The standard Gaussian generalized linear models (GLM) are not statistically valid for this type of data. Therefore, we implemented Binomial GLM with logit link function for the BayesB and Bayesian GBLUP genomic prediction methods. We compared these models with their standard Gaussian counterparts using two experimental data sets from plant breeding, one on female fertility in wheat and one on haploid induction in maize, as well as a simulated data set. With the aid of the simulated data referring to a bi-parental population of doubled haploid lines, we further investigated the influence of training set size (N), number of independent Bernoulli trials for trait evaluation (n i ) and genetic architecture of the trait on genomic prediction accuracies and abilities in general and on the relative performance of our models. For BayesB, we in addition implemented finite mixture Binomial GLM to account for overdispersion. We found that prediction accuracies increased with increasing N and n i . For the simulated and experimental data sets, we found Binomial GLM to be superior to Gaussian models for small n i , but that for large n i Gaussian models might be used as ad hoc approximations. We further show with simulated and real data sets that accounting for overdispersion in Binomial data can markedly increase the prediction accuracy.

Association between microsatellite markers and bovine tuberculosis in Chadian Zebu cattle

Bovine tuberculosis (BTB) is a considerable threat to livestock keepers and public health in many developing and underdeveloped countries. We investigated associations between 20 microsatellite markers and three phenotypes for BTB in a sample of the Chadian cattle population. The phenotypes measured were: 1) single intra-dermal comparative cervical tuberculin test (SICCT) performed on live animals, 2) presence of lesion post-slaughter, 3) a bacteriological tissue culture test for Mycobacterium bovis using the samples with observed lesions and 4) a predicted Bayesian model (BM) estimate of a true BTB disease status using all tested animals. All traits were recorded in binary form and as either 1 = presence or 0 = absence. A total of 224 animals for SICCT, lesion and BM traits and 96 animals with bacteriological culture test were genotyped. Generalised linear models were fitted to the binary BTB phenotypes that consisted of age (covariate), sex (2 levels), breed (2 levels) and markers (alleles: 5 - 14 levels) as explanatory variables and implemented in R using glm with a logit link function. The model was fitted for each marker, separately. Six out of 20 markers tested were significantly associated with at least one trait considered; these were ILSTS005, ILSTS006, TGLA227, BM2113 and CSRM66. Genomic regions around these markers may serve as a basis for further functional investigations. This is the first study to report association of microsatellite markers with bovine tuberculosis traits in African or Chadian cattle population.

Evidence Synthesis for Decision Making 2

We set out a generalized linear model framework for the synthesis of data from randomized controlled trials. A common model is described, taking the form of a linear regression for both fixed and random effects synthesis, which can be implemented with normal, binomial, Poisson, and multinomial data. The familiar logistic model for meta-analysis with binomial data is a generalized linear model with a logit link function, which is appropriate for probability outcomes. The same linear regression framework can be applied to continuous outcomes, rate models, competing risks, or ordered category outcomes by using other link functions, such as identity, log, complementary log-log, and probit link functions. The common core model for the linear predictor can be applied to pairwise meta-analysis, indirect comparisons, synthesis of multiarm trials, and mixed treatment comparisons, also known as network meta-analysis, without distinction. We take a Bayesian approach to estimation and provide WinBUGS program code for a Bayesian analysis using Markov chain Monte Carlo simulation. An advantage of this approach is that it is straightforward to extend to shared parameter models where different randomized controlled trials report outcomes in different formats but from a common underlying model. Use of the generalized linear model framework allows us to present a unified account of how models can be compared using the deviance information criterion and how goodness of fit can be assessed using the residual deviance. The approach is illustrated through a range of worked examples for commonly encountered evidence formats.