The Stata Journal: Promoting communications on statistics and Stata, 2005
We discuss methods for computing confidence intervals for predictions and discrete changes in pre... more We discuss methods for computing confidence intervals for predictions and discrete changes in predictions for regression models for categorical outcomes. The methods include endpoint transformation, the delta method, and bootstrapping. We also describe an update to prvalue and prgen from the SPost package, which adds the ability to compute confidence intervals. The article provides several examples that illustrate the application of these methods.
The multinomial logit model is perhaps the most commonly used regression model for nominal outcom... more The multinomial logit model is perhaps the most commonly used regression model for nominal outcomes in the social sciences. A concern raised by many researchers, however, is the assumption of the independence of irrelevant alternatives (IIA) that is implicit in the model. In this article, the authors undertake a series of Monte Carlo simulations to evaluate the three most commonly discussed tests of IIA. Results suggest that the size properties of the most common IIA tests depend on the data structure for the independent variables. These findings are consistent with an earlier impression that, even in well-specified models, IIA tests often reject the assumption when the alternatives seem distinct and often fail to reject IIA when the alternatives can reasonably be viewed as close substitutes. The authors conclude that tests of the IIA assumption that are based on the estimation of a restricted choice set are unsatisfactory for applied work.
International Journal of Public Opinion Research, 2009
The availability of cross-national survey data has grown exponentially in recent years. While muc... more The availability of cross-national survey data has grown exponentially in recent years. While much attention has been paid to increasing the comparability of indicators across countries, less has been done to increase the comparability of measurement models. This article examines the implicit assumptions of four different approaches to measurement modeling-summative scales, pooled exploratory factor analysis, multiple-group confirmatory factor analysis, and locally-conditioned factor analysis, and explores whether substantive conclusions in cross-national work can vary depending on the choice of measurement model. We find that results can vary by method and suggest that (i) the measurement modeling process itself be a critical part of cross-national research, and (ii) analysts be prepared to fully explain and defend measurement modeling decisions. A thorough understanding of the implicit assumptions of measurement modeling is required to avoid drawing conclusions that are little more than arbitrary. During the past two decades there has been a stunning growth in crossnational survey research, due in no small part to the increasing availability of cross-national survey data. Surveys such as the World Values Survey (WVS), European Social Survey (ESS), International Social Survey Programme (ISSP), and the various Barometers (now in Europe, Latin America, Africa, and Asia) exemplify this trend, covering domains as diverse as citizenship, environment, family and gender roles, leisure time and sports, national identity, religion, social inequality and social networks across a wide range of countries (see Heath, Fisher & Smith, 2005 for a review). The increasingly global reach of survey research has provided social scientists with new opportunities to pursue theory building and refinement through
The Stata Journal: Promoting communications on statistics and Stata, 2016
Hoaglin claims that regression coefficients are commonly, perhaps usually, misinterpreted. Citing... more Hoaglin claims that regression coefficients are commonly, perhaps usually, misinterpreted. Citing the preliminary edition of Tukey's classic Exploratory Data Analysis (1970, chap. 23), Hoaglin argues that the correct interpretation of a regression coefficient is that it "tells us how Y responds to change in X 2 after adjusting for simultaneous linear change in the other predictors in the data at hand". He contrasts this with what he views as the common misinterpretation of the coefficient as "the average change in Y for a 1-unit increase in X 2 when the other Xs are held constant". He asserts that this interpretation is incorrect because "[i]t does not accurately reflect how multiple regression works". We find that Hoaglin's characterization of common practice is often inaccurate and that his narrow view of proper interpretation is too limiting to fully exploit the potential of regression models. His article rehashes debates that were settled long ago, confuses the estimator of an effect with what is estimated, ignores modern approaches, and rejects a basic goal of applied research. In what follows, we assume the outcome variable is y, that x is the predictor whose effect is being interpreted, and that w represents one or more additional predictors in the model. Our examples are purposely simple, but the arguments would not change with more realistic specifications. Although our discussion is limited to linear regression, the ideas apply generally to the interpretation of marginal effects, such as ∂π(x)/∂x, in nonlinear models such as logit or probit. Before explaining why we disagree with key points in Hoaglin's argument, we note several points that we agree with: 1. The regression coefficient for x generally depends on the other predictors in the model.
This paper explores the possible measurable effects of mentors (major professors) on the subseque... more This paper explores the possible measurable effects of mentors (major professors) on the subsequent productivity of the mentor's students. Also asked is whether there are benefits to the productive scientist who acts as a mentor. Analysis is based on a population of male biochemists (N=66) who obtained their doctorates in 1957, 1958, 1962, and 1963. Various measures of departmental, mentor and/or student productivity and prestige were obtained. Sources included biographic information from American Men (and Women) of Science and productivity indications from the Science Citation Index and Chemical Abstracts. Results indicate no visible mentor effects, unless the postdoctoral context is one in which research productivity is encouraged. Other results indicate that the productivity of former students positively influences the research visibility of the mentor. (CS)
International Journal of Multiple Research Approaches, 2013
Made available courtesy of e-Content Management Pty Ltd.: http://pubs.econtentmanagement.com/loi/... more Made available courtesy of e-Content Management Pty Ltd.: http://pubs.econtentmanagement.com/loi/mra ***Reprinted with permission. No further reproduction is authorized without written permission from e-Content Management Pty Ltd. This version of the document is not the version of record. Figures and/or pictures may be missing from this format of the document.
A great deal of recent work in econometrics has focused on the development of tests to detect vio... more A great deal of recent work in econometrics has focused on the development of tests to detect violations of the assumptions of ordinary least squares regression. These tests are referred to collectively as specification tests. This article evaluates some important and computationally convenient specification tests for the normal regression model as applied to cross-sectional data. Because these tests achieve their optimal properties in large samples, their size and power in finite samples are of great interest and are evaluated with Monte Carlo simulations. Although the authors' experiments showed a tendency toward overrejection in some tests, their results suggest that specific variations of the RESET and information matrix tests behave quite well even in small samples. They conclude by proposing a strategy for the sequential application of specification tests.
As a consequence of the impact of Viagra on male sexual dysfunction, considerable attention is no... more As a consequence of the impact of Viagra on male sexual dysfunction, considerable attention is now being paid to sexual dysfunctions in women, which might respond to pharmacological treatment. Should women's sexual problems be conceptualized in the same way as men's? The objective of this study was to assess the prevalence of distress about sexuality among women, and examine the
Sexuality research focuses almost exclusively on individuals rather than couples, though ongoing ... more Sexuality research focuses almost exclusively on individuals rather than couples, though ongoing relationships are very important for most people and cultures. The present study was the first to examine sexual and relationship parameters of middle-aged and older couples in committed relationships of 1-51 years duration. Survey research was conducted in Brazil, Germany, Japan, Spain, and the U.S. targeting 200 men aged 40-70 and their female partners in each country, with 1,009 couples in the final sample. Key demographic, health, physical intimacy, sexual behavior, sexual function, and sexual history variables were used to model relationship happiness and sexual satisfaction. The median ages were 55 for men and 52 for women; median relationship duration was 25 years. Relationship satisfaction in men depended on health, physical intimacy, and sexual functioning, while in women only sexual functioning predicted relationship satisfaction. Models predicting sexual satisfaction included significant physical intimacy and sexual functioning for both genders and, for men, more frequent recent sexual activity and fewer lifetime partners. Longer relationship duration predicted greater relationship happiness and sexual satisfaction for men. However, women in relationships of 20 to 40 years were significantly less likely than men to report relationship happiness. Compared to men, women showed lower sexual satisfaction early in the relationship and greater sexual satisfaction later. Within the long-term committed relationship context, there were significant gender differences in correlates of sexual and relationship satisfaction, with sexual functioning a common predictor of both types of satisfaction and physical intimacy a more consistent and salient predictor for men.
In the United States, considerable attention has been directed to sexual behaviors of black and w... more In the United States, considerable attention has been directed to sexual behaviors of black and white adolescents, particularly age at first sexual experience and the prevalence of teenage pregnancies. More limited attention has been paid to comparing established sexual relationships in these two racial groups. In this study, we used a national probability sample to compare black (n = 251) and white (n = 544) American women, aged 20-65 years, who were in an established heterosexual relationship of at least 6 months duration. We focused on two aspects of their sexual well-being; how a woman evaluated (1) her sexual relationship and (2) her own sexuality. A range of possible determinants of sexual well-being, including demographic factors, physical and mental health, and aspects of the women's recent sexual experiences, were also assessed using Telephone-Audio-Computer-Assisted Self-Interviewing (T-ACASI). We found no significant difference between black and white women in their evaluation of their sexual relationships nor in the independent variables that were correlated with this evaluation. Black women, however, evaluated their own sexuality more positively than white women. In examining the correlates of this evaluation, a woman's rating of her own sexual attractiveness proved to be the strongest predictor, with black women rating themselves significantly more sexually attractive than did the white women. Overall, these findings were consistent with previous findings that, compared to white women, black women in the United States have higher self-esteem and tend towards more independence and individualism.
In the presence of heteroscedasticity, OLS estimates are unbiased, but the usual tests of signifi... more In the presence of heteroscedasticity, OLS estimates are unbiased, but the usual tests of significance are generally inappropriate and their use can lead to incorrect inferences. Tests based on a heteroscedasticity consistent covariance matrix (HCCM), however, are consistent even in the presence of heteroscedasticity of an unknown form. Most applications that use a HCCM appear to rely on the asymptotic version known as HC0. Our Monte Carlo simulations show that HC0 often results in incorrect inferences when N ≤ 250, while three relatively unknown, small sample versions of the HCCM, and especially a version known as HC3, work well even for N 's as small as 25. We recommend that: 1) data analysts should correct for heteroscedasticity using a HCCM whenever there is reason to suspect heteroscedasticity; 2) the decision to use a HCCM-based tests should not be determined by a screening test for heteroscedasticity; and 3) when N ≤ 250, the HCCM known as HC3 should be used. Since HC3 is simple to compute, we encourage authors of statistical software to add this estimator to their programs.
To extend our understanding of how social structures and social processes impact behavior, sociol... more To extend our understanding of how social structures and social processes impact behavior, sociologists have been challenged to incorporate the potential explanatory role of genetics in their models. Here, we draw propositions from three major understandings of illness causation offered by social theory-fundamental causes, social stress processes, and social safety net theories. We tailor hypotheses to the case of alcohol dependence, long considered a multifaceted problem, defying simple explanation and having both biological and social roots. After briefly reviewing current appeals for transdisciplinary research, we describe both sociological and genetic theories, and derive propositions expected under each and under a transdisciplinary theoretical frame. Analyses of a later wave of the preeminent medical science study, the Collaborative Study on the Genetics of Alcoholism (COGA), reveals a complex interplay of how the GABRA2 gene works with and against social structural factors to produce cases meeting DSM/ICD diagnoses. When both genetic and social factors are controlled, virtually equivalent effects of each remain; and, only modest evidence suggests that genetic influence works through social structural conditions and experiences. Further exploratory analyses using multiplicative terms reveal enhanced geneenvironment interactions: 1) women are largely unaffected in their risk for alcohol dependence by allele status at this candidate gene; 2) family support attenuates genetic influence; 3) childhood deprivation exacerbates genetic predispositions. We discuss how these findings lead us to consider the essential intradisciplinary tension in sociological theories (i.e., the role of proximal and distal influences in social processes). Overall, our findings point to the promise of theories blending social and genetic influences by focusing directly on dynamic, networked sequences that produce different pathways to health and illness.
Submissions to the STB, including submissions to the supporting files (programs, datasets, and he... more Submissions to the STB, including submissions to the supporting files (programs, datasets, and help files), are on a nonexclusive, free-use basis. In particular, the author grants to StataCorp the nonexclusive right to copyright and distribute the material in accordance with the Copyright Statement below. The author also grants to StataCorp the right to freely use the ideas, including communication of the ideas to other parties, even if the material is never published in the STB. Submissions should be addressed to the Editor. Submission guidelines can be obtained from either the editor or StataCorp. Copyright Statement. The Stata Technical Bulletin (STB) and the contents of the supporting files (programs, datasets, and help files) are copyright c by StataCorp. The contents of the supporting files (programs, datasets, and help files), may be copied or reproduced by any means whatsoever, in whole or in part, as long as any copy or reproduction includes attribution to both (1) the author and (2) the STB. The insertions appearing in the STB may be copied or reproduced as printed copies, in whole or in part, as long as any copy or reproduction includes attribution to both (1) the author and (2) the STB. Written permission must be obtained from Stata Corporation if you wish to make electronic copies of the insertions. Users of any of the software, ideas, data, or other materials published in the STB or the supporting files understand that such use is made without warranty of any kind, either by the STB, the author, or Stata Corporation. In particular, there is no warranty of fitness of purpose or merchantability, nor for special, incidental, or consequential damages such as loss of profits. The purpose of the STB is to promote free communication among Stata users.
Methods for group comparisons using predicted probabilities and marginal effects on probabilities... more Methods for group comparisons using predicted probabilities and marginal effects on probabilities are developed for regression models for binary outcomes. Unlike approaches based on the comparison of regression coefficients across groups, the methods we propose are unaffected by the scalar identification of the coefficients and are expressed in the natural metric of the outcome probability. While we develop our approach using binary logit with two groups, we consider how our interpretive framework can be used with a broad class of regression models and can be extended to any number of groups.
Self-reported discrimination is linked to diminished well-being, but the processes generating the... more Self-reported discrimination is linked to diminished well-being, but the processes generating these reports remain poorly understood. Employing the life course perspective, this paper examines the correspondence between expected age preferences for workers ...
The SPost user package (Long and Freese, 2006, Regression Models for Categorical Dependent Variab... more The SPost user package (Long and Freese, 2006, Regression Models for Categorical Dependent Variables Using Stata [Stata Press]) is a suite of postestimation commands to compute additional tests and effects representations for a variety of regression models. To facilitate and automate the task of tabulating results from SPost commands for inclusion in reports, publications, and presentations, we introduce tools to integrate
Uploads
Papers by Scott Long