Baseline Category Logit Model Research Articles

Abstract PO4-02-14: Parity and age first full term pregnancy affects the odds of different breast cancer subtypes defined by estrogen receptor and human epidermal growth factor receptor-2

Abstract Background: A rising trend in the incidence of estrogen receptor (ER)-positive subtype over time is observed in several high-income countries, while the incidence of ER-negative subtype is decreasing in some areas. It has been found that decreased parity and late age at 1st full-term pregnancy (FTP1) can increase the risk of breast cancer (BC). This study aims at investigating the association between two reproductive factors, parity and FTP1, and the odds of BC subtypes in premenopausal and postmenopausal patients. Patients and method: This study included patients diagnosed with early-stage BC after the age of 34, who were treated in UZ Leuven between January 2000 and November 2020. The patients were classified into four subtypes defined by ER and human epidermal growth factor receptor-2 (HER2). ER positivity was defined as >1% of cells staining positive while HER2 positivity was based on fluorescence in situ hybridization (FISH) testing of samples with immunohistochemistry score of 2+ or 3+. Baseline-category logit models with subtype as the response (ER+/HER2- as the baseline) and adjusted for age at diagnosis and BMI, were performed for premenopausal and postmenopausal patients separately. First, we exclude nulliparous patients. Parity (1-2 or ≥3) and FTP1 (continuous) were included in the models as the covariates of interest in the first model. To evaluate the effects of reproductive patterns compared to nulliparity, a new variable reproductive pattern was created by taking intersection of parity groups and FTP1 groups (< 27, ≥27). Reproductive pattern was set as the covariate of interest in the second model. Result: After imputing the missing values, we included 9606 patients in the study. 8125 of them were parous. In parous patients, high FTP1 was associated with low odds of ER-/HER2- BC relative to ER+/HER2- BC for both premenopausal and postmenopausal patients (table 1). In postmenopausal patients, low parity was associated with low odds of ER-/HER2- BC and high odds of ER+/HER2+ BC relative to ER+/HER2- BC. There was no evidence that parity affects the odds of subtypes in premenopausal patients. When investigating the effect of different reproductive patterns compared to nulliparity on the odds of subtypes, there was no evidence that reproductive patterns will affect the odds of subtypes in premenopausal patients. Postmenopausal patients who had FTP1 before 27 years old and had at least three children tended to have higher odds of ER-/HER2- BC relative to ER+/HER2- BC compared to nulliparous patients. Conclusion: In parous patients, increased FTP1 was associated with reduced odds of ER-/HER2- relative to ER+/HER2- BC. In postmenopausal patients, patients with at least three children tended to have low odds of ER-/HER2- BC and high odds of ER+/HER2+ BC relative to ER+/HER2- BC. When compared to nulliparity, the combination of having at least three children and having the first child before 27 years old increased the odds of ER-/HER2- BC relative to ER+/HER2- BC in postmenopausal patients. Table 1 Effect of FTP1, parity on subtype odds Menopause Subtype Effect Odds ratio (95%CI) p-value pre ER-/HER2- FTP1 (continuous) 0.960 (0.935, 0.985) 0.0021 post 0.969 (0.948, 0.990) 0.0047 pre ER-/HER2+ 0.992 (0.952, 1.034) 0.7166 post 1.005 (0.977, 1.034) 0.7081 pre ER+/HER2+ 0.976 (0.949, 1.004) 0.0962 post 1.015 (0.992, 1.039) 0.1969 pre ER-/HER2- Parity 1-2 vs ≥3 1.194 (0.899, 1.585) 0.2200 post 0.784 (0.648, 0.949) 0.0124 pre ER-/HER2+ 1.152 (0.723, 1.834) 0.5521 post 0.896 (0.684, 1.173) 0.4228 pre ER+/HER2+ 1.080 (0.795, 1.469) 0.6220 post 1.300 (1.022, 1.654) 0.0329 The baseline category is ER+/HER2- subtype Effect of age at first full-term pregnancy and parity on subtype odds Citation Format: Zhao Liu, Jiumeng Zhang, Karen Van Baelen, Maja Vangoitsenhoven, Hans Wildiers, Adelheid Soubry, Patrick Neven. Parity and age first full term pregnancy affects the odds of different breast cancer subtypes defined by estrogen receptor and human epidermal growth factor receptor-2 [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO4-02-14.

Assessment of the risk factors associated with type 2 diabetes and prediabetes mellitus: A national survey in Vietnam.

This study aims to estimating the prevalence of type 2 diabetes and prediabetes among adult from 30 to 69 years old and assess the association of risk factor with the conditions. A total of 5244 aged 30 to 69 years old were participated in this cross-sectional study, using nationally representative sampling frame. All participants were taking blood sample to measure fasting blood glucose level and 2-hour postload oral glucose tolerance test by National Hospital of Endocrinology, Vietnam. Multinomial logistic regressions with baseline-category logit models were conducted to identify factors associated with diabetes and prediabetes among respondents. The prediabetes prevalence was in 17.9% and diabetes in 7.3%. Patients who were male (reference group vs female OR = 0.79; 95% CI: 0.64, 0.97), in the 50 to 59 years old group (OR = 1.60; 95% CI: 1.28, 2.00), have hypertension and WHR risk have higher prevalence to have prediabetes (OR = 1.31; 95% CI: 1.12, 1.53; OR = 1.37; 95% CI: 1.11, 1.70, respectively). Male patients (reference group vs female OR = 0.62; 95% CI: 0.45, 0.84), patients who were in 40 to 49; 50 to 59; 60 to 69 years old, those who were housewife (OR = 2.17; 95% CI: 1.43, 3.28; OR = 2.85; 95% CI: 1.91, 4.27; OR = 3.12; 95% CI: 2.08, 4.69; OR = 1.71; 95% CI: 1.22, 2.40, respectively). Diabetes have significant associated with participants have hypertension (OR = 1.19; 95% CI: 1.72, 2.70). The common factor directly related to prediabetes and diabetes in both genders is age. Other factors directly associated with prediabetes and diabetes include BMI, WHR, hypertension, educational level, and job.

Holling meets habitat selection: functional response of large herbivores revisited

BackgroundHolling (Can Entomol 91(5):293–320, 1959) was the first to describe a functional response between a predator’s consumption-rate and the density of its prey. The same concept can be applied to the habitat selection of herbivores, specifically, the change in relative habitat use with the change in habitat availability. Functional responses in habitat selection at a home-range scale have been reported for several large herbivores. However, a link to Holling’s original functional response types has never been drawn, although it could replace the current phenomenological view with a more mechanistically based understanding of functional responses.MethodsIn this study, discrete choice models were implemented as mixed-effects baseline-category logit models to analyze the variation in habitat selection of a large herbivore at seasonal and diurnal scales. Thus, changes in the use of land cover types with respect to their availability were investigated by monitoring 11 land cover types commonly used by roe deer (Capreolus capreolus) in the Bavarian Forest National Park, Germany. Functional response curves were then fitted using Holling’s formulas.ResultsStrong evidence of non-linear functional responses was obtained for almost all of the examined land cover types. The shape of the functional response curves varied depending on the season, the time of day, and in some cases between sexes. These responses could be referenced to Holling’s types, with a predominance of type II.ConclusionsOur results indicate that Holling’s types can be applied to describe general patterns of the habitat selection behavior of herbivores. Functional responses in habitat selection may occur in situations requiring a trade-off in the selection of land cover types offering different resources, such as due to the temporally varying physiological needs of herbivores. Moreover, two associated parameters defining the curves (prey density and predation rate) can aid in the identification of temporal variations and in determinations of the strength of the cost-benefit ratio for a specific land cover type. Application of our novel approach, using Holling’s equations to describe functional responses in the habitat selection of herbivores, will allow the assignment of general land cover attraction values, independent of availability, thus facilitating the identification of suitable habitats.

Trends in shaken baby syndrome diagnosis codes among young children hospitalized for abuse

ObjectiveTo investigate national trends of SBS diagnosis codes and how trends varied among patient and hospital characteristics.MethodsWe examined possible SBS, confirmed SBS, and non-SBS abuse diagnosis codes among children age three and younger who were hospitalized for abuse between 1998 and 2014 using a secondary analysis of the National Inpatient Sample, the largest US all-payer inpatient care database (N = 66,854). A baseline category logit model was used based on a quasi-likelihood approach (QIC) with an independent working correlation structure.ResultsThe rate (per 100,000 census population of children age 3 and younger) of confirmed and possible SBS diagnosis codes was 5.4 (± 0.3) between 1998 and 2014, whereas the rate of non-SBS abuse was 19.6 (± 1.0). The rate of confirmed SBS diagnosis codes increased from 3.8 (± 0.3) in 1998 to 5.1 (± 0.9) in 2005, and decreased to 1.3 (± 0.2) in 2014. Possible SBS diagnosis codes were 0.6 (± 0.2) in 1998, increasing to 2.4 (± 0.4) in 2014. Confirmed SBS diagnosis codes have declined since 2002, while possible SBS diagnosis codes have increased. All abuse types were more frequent among infants, males, children from low-income homes, and urban teaching hospitals.ConclusionsWe investigated seventeen-year trends of SBS diagnosis codes among young children hospitalized for abuse. The discrepancy between trends in possible and confirmed SBS diagnosis codes suggests differences in norms for utilizing SBS diagnosis codes, which has implications for which hospital admissions are coded as AHT. Future research should investigate processes for using SBS diagnosis codes and whether all codes associated with abusive head injuries in young children are classified as AHT. Our findings also highlight the relativity defining and applying SBS diagnosis codes to children admitted to the hospital for shaking injuries. Medical professionals find utility in using SBS diagnosis codes, though may be more apt to apply codes related to possible SBS diagnosis codes in children presenting with abusive head injuries. Clarifying norms for SBS diagnosis codes and refining definitions for AHT diagnosis will ensure that young children presenting with, and coded for, abusive head injuries are included in overall counts of AHT based on secondary data of diagnosis codes. This baseline data, an essential component of child abuse surveillance, will enable ongoing efforts to track, prevent, and reduce child abuse.

A note on the three methods of item analysis

This paper contrasts three methods of item analysis for multiple-choice items based on classical test theory, generalized linear modeling, and item response theory. Illustrations of the methods are presented with contrived and real data. Specifically, the methods, respectively, use a cross-classification table under classical test theory, a new baseline-category logit model under generalized linear modeling, and a multiple-choice model under item response theory. Advantages and disadvantages of each method are discussed.

Women’s Autonomy in Household Purchasing Decision Making in Sri Lanka: An Application of Multivariate Baseline-Category Logit Model

As many researches on gender and household purchasing decision making are sociological as well as descriptive, this attempt tries to fill the gap with applying of statistical model in to the same context. This is basically an inductive approach that generates data, applies joint modeling and moves forward in sociological interpretation. The data has been generated in an Island-wide survey consisting 28,800 household units in Sri Lanka in 2016. Generalized linear mixed approach was applied to baseline-category logit models to identify the factors associated with women’s autonomy in decision making on daily household purchases and major household purchases. This research comes up to the conclusions that women’s age, education status of the women and spouses, number of children, residence area and economic status in the family emerged as important factors associated with women’s autonomy in decision making. Increased age, well paid employment and having many living children are all positively associated with women’s autonomy in decision making in both daily household purchases and major household purchases. Women from rural and estate areas are less likely be autonomous in decision making than the women from urban areas. Highly educated women are more likely to make decisions jointly while the women who less attended formal education are further likely to make decisions unaccompanied.

Eye movement control in Turkish sentence reading.

Reading requires the assembly of cognitive processes across a wide spectrum from low-level visual perception to high-level discourse comprehension. One approach of unravelling the dynamics associated with these processes is to determine how eye movements are influenced by the characteristics of the text, in particular which features of the words within the perceptual span maximise the information intake due to foveal, spillover, parafoveal, and predictive processing. One way to test the generalisability of current proposals of such distributed processing is to examine them across different languages. For Turkish, an agglutinative language with a shallow orthography-phonology mapping, we replicate the well-known canonical main effects of frequency and predictability of the fixated word as well as effects of incoming saccade amplitude and fixation location within the word on single-fixation durations with data from 35 adults reading 120 nine-word sentences. Evidence for previously reported effects of the characteristics of neighbouring words and interactions was mixed. There was no evidence for the expected Turkish-specific morphological effect of the number of inflectional suffixes on single-fixation durations. To control for word-selection bias associated with single-fixation durations, we also tested effects on word skipping, single-fixation, and multiple-fixation cases with a base-line category logit model, assuming an increase of difficulty for an increase in the number of fixations. With this model, significant effects of word characteristics and number of inflectional suffixes of foveal word on probabilities of the number of fixations were observed, while the effects of the characteristics of neighbouring words and interactions were mixed.

Sex Differences and Psychological Factors Associated with General Health Examinations Participation: Results from a Vietnamese Cross-Section Dataset

This study focuses on the association of sex differences and psychological factors with periodic general health examination (GHE) behaviors. We conducted a survey in Hanoi and the surrounding areas, collecting 2068 valid observations; the cross-section dataset was then analyzed using the baseline category logit model. The study shows that most people are afraid of discovering diseases through general health examinations (76.64%), and the fear of illness detection appears to be stronger for females than for males (β1(male) = −0.409, p < 0.001). People whose friends/relatives have experienced prolonged treatment tend to show more hesitation in participating in physical check-ups (β2 = 0.221, p < 0.05). On the ideal frequency of GHEs, 90% of the participants agree on once or twice a year. The probability of considering a certain period of time as an appropriate frequency for GHEs changes in accordance with the last doctor visit (low probability of a health examination every 18 months) and one’s fear of potential health problems post-checkup (no fear raises probability of viewing a health examination every 6 months by 9–13 percentage points). The results add to the literature on periodic GHE in particular and on preventive health behaviors in general.

Factors Associated with the Regularity of Physical Exercises as a Means of Improving the Public Health System in Vietnam

Being ranked among the most sedentary countries, Vietnam’s social public health is challenged by the rising number of overweight people. This study aims to evaluate factors associated with the regularity of exercise and sports (EAS) among Vietnamese people living in the capital city of Hanoi, using data collected from a randomized survey involving 2068 individuals conducted in 2016. Physical exercises and daily sports are considered a major means for improving the Vietnamese social public health system by the government, families, and individuals. Applying the baseline-category logit model, the study analyzed two groups of factors associated with EAS regularity: (i) physiological factors (sex, body mass index (BMI)) and (ii) external factors (education, health communication, medical practice at home). Females with a university education or higher usually exercise less than those with lower education, while the opposite is true for males. The study also shows that those with a higher BMI tend to report higher activity levels. Additionally, improved health communication systems and regular health check-ups at home are also associated with more frequent EAS activities. These results, albeit limited to only one location in Vietnam, provide a basis for making targeted policies that promote a more active lifestyle. This, in turn, could help the country realize the goal of improving the average height of the population and reducing the incidents of non-communicable diseases.

Longitudinal model for categorical data applied in an agriculture experiment about elephant grass

Experiments where the response is a categorical variable are usually carried out in many fields such as agriculture. In addition, in some situations this response has three or more levels without an order between them characterizing a multinomial (nominal) response. Statistical models for scenarios where the observations of a nominal response can be considered independent have an extensive literature, such as the baseline-category logit models. However, situations where this assumption is violated (as in longitudinal studies) require specific models that take into consideration the dependence between observations. In this paper, a fairly new extension of the generalized estimating equations is applied to analyze an experiment carried out to investigate the type of vegetation observed in an elephant grass pasture, according to some management conditions over time. This extension uses local odds ratios to explain the dependence among the categories of the outcome over the repeated measurements. Two different structures were compared to describe this dependence, and the Wald test was used to select the significant variables. Further, we built confidence intervals for the predicted probabilities of occurrence of each category and assessed the results comparing observed/predicted values and using the diagnostic analysis. The results allowed to conclude that there are various significant effects for treatments and for time. The structure of local odds ratio also proved as a good way to describe the dependence between categorical responses over time.

How important is perception of safety to park use? A four-city survey.

To determine individual- and park-related characteristics associated with adults visiting their closest neighborhood park and involvement in sports, walking, and sedentary activities. Cross-sectional study. Neighborhoods surrounding 24 parks in four United States metropolitan areas. Adults (N = 3,815) (25% African American, 12% Hispanic, and 56% Non-Hispanic White) living within .5 mile of one of 24 public parks. Anonymous surveys and park observations. Chi-square statistics and logistic regression including individual and park characteristics. Thirty-seven percent reported never having visited their park; 16% reported using it for regular exercise. Respondents who perceived the parks as safe or very safe had 4.6 times the odds of visiting the parks. Active sports participation was associated with being male, ages 18 to 24, and of African American heritage. Walking was associated with being female, age 47+, and reporting fair or poor health. Females, Hispanics, and those reporting fair/poor health were more likely to report sedentary activities. Park characteristics associated with park use included fewer physical incivilities, closer proximity to homes, more facilities, and organized activities. Perceptions of park safety were strongly associated with ever having visited a respective park, while park characteristics appeared to influence the types of activities performed. Increasing the number of organized activities and offering a variety of different facilities may encourage park use.

Intimate association of visceral obesity with non‐alcoholic fatty liver disease in healthy Asians: A case‐control study

To identify factors associated with non-alcoholic fatty liver disease (NAFLD) in healthy Asian subjects. A propensity score-matched case-control study was designed. To investigate the effects of demographic and clinical factors on the presence of NAFLD, a baseline-category logit model was used. Potential living liver donors with no hepatic steatosis (< 5%: n = 1353, group 1) were considered the baseline category, and subjects with mild (5-33%: n = 724, group 2) and moderate/severe (> 33%: n = 116, group 3) hepatic steatosis were defined as cases. Age and gender were matched between cases and controls, which resulted in 83 matched subjects in each of the three groups. The area of abdominal (visceral and subcutaneous) fat was directly measured in all subjects by unenhanced computed tomography. Serum aspartate aminotransferase, alanine aminotransferase (ALT), gamma-glutamyltranspeptidase, total cholesterol and triglyceride levels, and visceral fat amount were directly correlated with the grade of hepatic steatosis, and high-density lipoprotein cholesterol levels were inversely correlated with it (all P values < 0.05). In a multivariate model, visceral fat amount was significantly correlated with both mild (group 2) and moderate to severe (group 3) NAFLD, with respective odds ratios (ORs) of 1.03 relative to group 1 (Ps < 0.05). Body mass index (BMI), ALT, and subcutaneous fat were significant predictors of only moderate to severe NAFLD (ORs of 0.54, 1.20, and 1.02, respectively, for group 3 vs group 1; Ps < 0.05). Our results indicate that visceral adiposity makes non-obese subjects more susceptible to NAFLD, compared with subcutaneous fat and BMI.

Multinomial Logit Modeling of Factors Associated With Multiple Sexual Partners from the Kenya Aids Indicator Survey 2007

The number of lifetime sex partners of an individual has an important effect on Human Immunodeficiency Virus (HIV) status of an individual; hence modeling multiple sexual partnerships is an essential component of any analysis of HIV outcome. Multiple sexual partnerships are associated with greater risk of HIV, Sexually Transmitted infections (STIs) and intimate partner violence. This research project presents a general approach for modeling logit of clustered (correlated) ordinal and nominal responses using polytomous data from the Kenya AIDS Indicator Survey 2007 (NASCOP 2010). We review multinomial logit models as generalized linear models. The model is applied to HIV prevalence data among men and women in Kenya, derived from the Kenya AIDS Indicator Survey 2007 (KAIS). We generalize logistic regression to handle multinomial response variables, with separate models for nominal and ordinal cases. When modeling a nominal response variable we are interested in finding if certain predictors have an effect on the probabilities. The baseline category logit model, models the odds of being in one category relative to being in a designated category (last category), for all pairs of categories. It is used for nominal responses. A maximum likelihood estimation (MLE) approach is used for baseline category logit model. To model an ordinal response variable one models the cumulative response probabilities or cumulative odds. The cumulative logit model is used when the response of an individual unit is restricted to one of a finite number of ordinal values. This study shows the practicality of multinomial logit model in analyzing epidemiological data. Other studies have found education to be associated with multiple sexual partners. In this study, we observed that multiple sexual partners is not related to education. Other covariates like Gender, Place of residence, sexually active individuals for the past 12 months and marital status were found to be associated with multiple sexual partners. Individuals that are sexually active for the past 12 months were found to be ten times more likely to have multiple sexual partners compared to those that are not. After controlling for all other factors, the odds of male to female having multiple sexual partners doubled to 7.6 meaning male are almost 8 times likely to have multiple sexual partners compared to female. Partner testing or couples testing is a main strategy of national testing initiatives in Kenya. Respondents are encouraged to learn their test results with their partner.

RPackagemultgee: A Generalized Estimating Equations Solver for Multinomial Responses

The R package multgee implements the local odds ratios generalized estimating equations (GEE) approach proposed by Touloumis, Agresti, and Kateri (2013), a GEE approach for correlated multinomial responses that circumvents theoretical and practical limitations of the GEE method. A main strength of multgee is that it provides GEE routines for both ordinal (ordLORgee) and nominal (nomLORgee) responses, while relevant other softwares in R and SAS are restricted to ordinal responses under a marginal cumulative link model specification. In addition, multgee offers a marginal adjacent categories logit model for ordinal responses and a marginal baseline category logit model for nominal responses. Further, utility functions are available to ease the local odds ratios structure selection (intrinsic.pars) and to perform a Wald type goodness-of-fit test between two nested GEE models (waldts). We demonstrate the application of multgee through a clinical trial with clustered ordinal multinomial responses.

English

Self- reported health status is the most commonly used measures of subjective and global measure of health because it is simple, economical and easy to administer. The objective of the study is to compare the performance of logistic regression models having multinomial response and identify the factors affecting health status of adolescents. Based on two stage sampling technique 2084 adolescents were interviewed to study the health status of teenagers in Jimma zone. In this article, we reviewed the most important logistic regression model and common approaches used to verify goodness-of-fit, using software R. We performed formal as well as graphical analyses to compare ordinal logistic regression models using data sets of health status. The results obtained from both baseline category logit model and ordinal logistic regression showed that sex of adolescents, source of drinking water and educational status significantly affect health status of teenagers. It was also found that a cumulative logit model containing these predictors provided the best description of the dataset among baseline category logit model, adjacent category logit model and continuation ratio model. &nbsp; Key words:&nbsp;Adolescents&rsquo; health status, multinomial logistic regression and ordinal logistic regression models, model comparisons using Akakie information criteria (AIC), goodness of fit.

Vitamin D Status and Demographic and Lifestyle Determinants among Adults in the United States (NHANES 2001-2006)

This study looked at risk factors associated with vitamin D levels in the body among a representative sample of adults in the U.S., NHANES III (2001-2006) data were used to assess the relationship between several demographic and health risk factors and vitamin D levels in the body. The Baseline-Category Logit Model was used to test the association between vitamin D level and the potential risk factors age, education, ethnicity, poverty status, physical activity, smoking, alcohol, obesity, diabetes and total cholesterol with both genders. Vitamin D insufficiency and deficiency were significantly associated with age, race, education, physical activity, obesity, diabetes and total cholesterol level for both genders. Almost half of the adults sampled in these data had vitamin D levels lower than the recommended limits, with the highest frequency among the younger groups. Determining an individual's vitamin D level is very difficult without proper clinical testing. Many of those who have low vitamin D levels are unaware. With such a high prevalence of individuals with low vitamin D levels in the U.S. and a better understanding of characteristics associated with these lower levels, increased education and prevention efforts should be focused toward those with higher risk characteristics.

원인균별 식중독 발생 건수 예측

이 논문에서는 우리나라에서 발생하는 원인균별 식중독 발생건수를 예측하는 방법을 제안한다. 우리나라에서 보고되는 주별 식중독 발생 건수를 원인균로 나누면 자료에 많은 0의 관측값이 포함되어 있으며 식중독 발생 간에 종속성을 가진다. 이 현상을 모형화하기 위해 이 논문에서는 전체 식중독 건수를 자기회귀모형으로 예측하고 원인균별 식중독 발생 확률을 다범주 로짓모형으로 추정한다. 예측된 식중독 건수와 추정된 원인균별 식중독 발생 확률을 곱하여 원인균별 식중독 발생건수를 예측한다. 제안된 방법의 타당성을 확인하기 위해 평균제곱오차와 평균절대편차를 이용하여 제안 방법과 영과잉모형을 비교해 본다. This paper proposes a method to predict the number of foodborne disease outbreaks by microbes. The weekly data of food poisoning occurrences by microbes in Korea contain many zero-valued observations and have dependency between outbreaks. In order to model both phenomena, the number of food poisonings is predicted by an autoregressive model and the probabilities of food poisoning occurrences by microbes (given the total of food poisonings) are estimated by the baseline category logit model. The predicted number of foodborne disease outbreaks by a microbe is obtained by multiplying the predicted number of foodborne disease outbreaks and the estimated probability of the food poisoning by the corresponding microbe. The mean squared error and the mean absolute value error are evaluated to compare the performances of the proposed method and the zero-inflated model.

Conditional pseudolikelihood methods for clustered ordinal, multinomial, or count outcomes with complex survey data

In order to adjust individual-level covariate effects for confounding due to unmeasured neighborhood characteristics, we have recently developed conditional pseudolikelihood methods to estimate the parameters of a proportional odds model for clustered ordinal outcomes with complex survey data. The methods require sampling design joint probabilities for each within-neighborhood pair. In the present article, we develop a similar methodology for a baseline category logit model for clustered multinomial outcomes and for a loglinear model for clustered count outcomes. All of the estimators and asymptotic sampling distributions we present can be conveniently computed using standard logistic regression software for complex survey data, such as sas proc surveylogistic. We demonstrate validity of the methods theoretically and also empirically by using simulations. We apply the new method for clustered multinomial outcomes to data from the 2008 Florida Behavioral Risk Factor Surveillance System survey in order to investigate disparities in frequency of dental cleaning both unadjusted and adjusted for confounding by neighborhood.

The Mixed Effects Trend Vector Model

Maximum likelihood estimation of mixed effect baseline category logit models for multinomial longitudinal data can be prohibitive due to the integral dimension of the random effects distribution. We propose to use multidimensional unfolding methodology to reduce the dimensionality of the problem. As a by-product, readily interpretable graphical displays representing change are obtained. The methodology can be applied to both nominal and ordinal response variables. Relationships to standard statistical models for multinomial data are presented. Several empirical examples are given to show the merits of the proposed modeling framework.

International determinants of private school attendance

The current study uses Programme for International Student Assessment (PISA) 2006 data to investigate international determinants of private school attendance. In particular, we seek to understand whether student achievement and home background factors such as socioeconomic status (SES) or motivational and goal-oriented factors are more predictive of private school attendance – in line with the concept of “cream-skimming’. Using baseline category logit models for nominal responses, our findings indicate that SES, rather than achievement or attitude, is the strongest, most consistent international predictor of attending a school managed or funded in large part by the private sector. Instead of traditional cream-skimming, our analysis suggests that a sorting effect of another kind exists – private schools (managed or funded) tend to lure students from better socioeconomic backgrounds, while public schools tend to attract the most engaged students.