1 item has been added to your cart.

Previous research on surgical outcomes (including morbidity, mortality, and subsequent healthcare utilization) has highlighted provider technique and experience, center volume-outcomes relationships, and geographical disparities in the quality of surgical care as important applications of clustered data analysis. In regression models, the nonindependence of outcomes within each cluster may be handled through cluster–robust standard errors or introduction of cluster-level fixed or random effects. However, clinical studies rarely articulate and occasionally misinterpret the rationale for applying these methods. I review recent literature on surgical outcomes to describe how the choice of approach may be influenced by the intended comparison among clusters, theoretical expectation of specific cluster-level factors influencing patient outcomes, and clinical importance of residual variation among clusters. I then present an example from transplant surgery where the primary contribution of a mixed-effects model is made by interpreting residual county-level variation in posttransplant survival.

stratum is poorly covered in its own way. I describe a new command to make what we call organ pipe plots to visualize heterogeneity in binary outcomes in clustered data. The plots were conceived for vaccination coverage surveys, but they are helpful in a wide variety of contexts. Imagine a survey where only 50% of sampled children are found to be vaccinated. Different programmatic responses would be appropriate if the vaccinated include all the children in half the clusters versus half the children in all the clusters. These plots have been used to identify neighborhoods that were surreptitiously and intentionally skipped over during vaccination campaigns. The talk will demonstrate the command and discuss similarities with Pareto plots from quality control and a visual connection to the intracluster correlation coefficient (ICC). Note that the ICC shares a connection to anarcho-pacifistic ideas in Tolstoy’s later novels: many students mention them … but few can describe them clearly.

Clinical markers and occupation were evaluated in 4 quartiles of exposure—0–2 μg/dL, 2–5 μg/dL, 5–10 μg/dL, and 10 μg/dL and over—to examine how the markers and various occupations manifested in the quartiles. Linear regression determined associations, and binary logistic regression predicted the likelihood of elevated clinical markers using binary degrees of exposure set at 2 μg/dL, 5 μg/dL, and 10 μg/dL. Clinical markers, and how they manifested between exposed and less exposed occupations, were explored in addition to how duration of exposure altered these clinical markers. In regression analysis, gamma-glutamyl transferase (GGT), total bilirubin, and alkaline phosphatase (ALP) were positively and significantly associated with blood lead level (BLL). Using binary logistic regression models, at the binary 2 μg/dL level, ALP and GGT were more likely to be elevated in those exposed. At 5 μg/dL level, it was ALP and GGT that were more likely to be elevated in those exposed, whereas at the 10 μg/dL level, it was GGT that was more likely to be elevated in those exposed. In the occupational analysis, aspartate aminotransferase (AST), alanine aminotransferase (ALT), GGT, and ALP showed differences between populations in the exposed and less exposed occupations. Regarding agriculture, forestry and fishing, duration of exposure altered AST, ALP, and total bilirubin significantly (p < 0.05), while ALT and GGT were altered moderately significantly (p < 0.10). With mining, duration of exposure altered AST and GGT moderately significantly, whereas in construction, duration in occupation altered AST and GGT significantly and total bilirubin moderately significantly. The study findings are evidence of occupational exposure to lead playing a significant role in initiating and promoting adverse hepatobiliary clinical outcomes in United States adults.

analytical methods that are commonly found in Bayesian analyses. However, almost every undergraduate is familiar with the t test. This presentation will use Stata's bayesmh command to perform a two-sample independent t test. We will discuss the advantages of using a Bayesian approach to perform t test-type analyses and compare the output or results with the traditional frequentist t test.

figures, footnotes, and paragraphs directly in Word or pdf files. The new command, putdocx/putpdf, serves as a one-stop-shop tool for transforming your Stata codes into Word or pdf files. This presentation will show you how to generate analysis tables, figures, and discussion or summary paragraphs directly in Word or pdf format. Plus, instead of manually updating the new numbers in your tables, figures, summary paragraphs, or footnotes when periodic updates are required, all you must do is refresh the dataset and run your existing do-file of putdocx/putpdf and call it to see the instantly updated results directly in the Word/pdf file. This can be done in one click. More specifically, below is a list of formatting and analysis results to be shown out of putdocx/putpdf and output directly in a Word or pdf file: 1. Paragraphs with statistics in them 2. Figures 3. Tables • descriptive summary table • regression table • logistic regression table • survival analysis table, etc. 4. Automation of exporting 5. Combination of several .docx files into one summary report.

probabilities and observed outcome rates. Significant deviations from the perfect calibration can be spotted on the graph. The graphical approach is paired to a statistical test, synthesizing the calibration assessment in a standard hypothesis testing framework. We present the calibrationbelt command, which implements the calibration belt and its associated test in Stata.

Parameters in these subject-specific nonlinear functions often have natural physical interpretations, and observations within the same subject are correlated. Subjects may be nested within higher-level groups, giving rise to nonlinear multilevel models, also known as nonlinear mixed-effects or hierarchical models. The new Stata 15 command menl allows you to fit nonlinear mixed-effects models, in which fixed and random effects may enter the model nonlinearly at different levels of hierarchy. In this talk, I will show you how to fit nonlinear mixed-effects models that contain random intercepts and slopes at different grouping levels with different covariance structures for both the random effects and the within-subject errors. I will also discuss parameter interpretation and highlight postestimation capabilities.

are fit via maximum likelihood and that also account for endogenous covariates, sample selection, and nonrandom treatment assignment. These models can be used when you are worried about bias due to unmeasured confounding, trials with informative dropout, outcomes that are missing not at random, selection on unobservables, and more. ERMs provide a unifying framework for handling these complications individually or in combination. I will briefly review the types of complications that ERMs can address. I will work through examples that demonstrate several of these complications and show some inferences we can make despite those complications.

SIMPLE was illustrated in the context of two-stage residual inclusion (2SRI) estimation (Terza et al., 2008). Stata/Mata implementations of SIMPLE for 2SRI estimators are detailed in Terza (2017a and b). Terza (2016b) develops a variant of SIMPLE for calculating the standard errors of two-stage marginal effects estimators (2SME). Generally applicable Stata/Mata implementation of SIMPLE for 2SME is detailed in Terza (2017c) and compared with results from the Stata margins command (for the subset of cases in which the margins command is available). Although SIMPLE substantially reduces the analytic and coding burden imposed by the conventional formulation, it still requires the derivation and coding of key partial derivatives that may prove daunting for some model specifications. In this presentation, I detail how such analytic demands and coding requirements are virtually eliminated via the use of the Mata deriv command. I will discuss illustrations in the 2SRI and 2SME contexts.

References:

Terza, J., A. Basu, and P. Rathouz (2008). Two-stage residual inclusion estimation. Addressing endogeneity in health econometric modeling. Journal of Health Economics 27: 531–543.

Terza, J.V. (2016a). Simpler standard errors for two-stage optimization estimators. Stata Journal 16: 368–385.

Terza, J.V. (2016b). Inference using sample means of parametric nonlinear data transformations. Health Services Research 51: 1109–1113.

Terza, J.V. (2017a). Two-stage residual inclusion estimation: A practitioners guide to Stata implementation. Stata Journal 17: 916–938.

Terza, J.V. (2017b). Two-stage residual inclusion estimation in health services research and health economics. Health Services Research, forthcoming, DOI: 10.1111/1475-6773.12714.

Terza, J.V. (2017c). Causal effect estimation and inference using Stata. Stata Journal 17: 939–961.

However, in the aggregate, we know there is a tendency to see such deposits become increasingly withdrawn as these accounts get older. Having an estimate of the lifetime of such deposit accounts is an important ingredient for calculating their present value. We show how to model the average lifetime based, first, on estimating the decay rate of deposit balances using Stata's nl command and, second, on calculating average lifetime based on the decay rate.

visual dimensions for exploratory and expositive purposes. While these tools are powerful on their own, they do not attempt to solve the most significant resource constraint we all face. The eda command is designed to address this time constraint by automating the creation of all the univariate and bivariate data visualizations and summary statistics tables in a dataset. Users can specify categorical and continuous variables manually, provide their own rules based on the number of unique values, or allow eda to use its own defaults, and eda will apply the necessary logic to graph and describe the data available. The command is designed to produce the maximum amount of output by default, so a single line of code can easily produce a document providing substantial insight into your data.

Methods: We applied Bland–Altman analysis (baplot) and ordinary least-products (OLP) regression (manually) in three simulated pairs of samples (N=100). In OLP, values of y and x are used in the major axis regression analysis, but then intercept and slope are back-transformed by dividing them by (). Fixed bias was defined if 95% confidence interval (CI) of the intercept does not include 0. Proportional bias was defined if 95%CI of the slope does not include 1. Results: Using baplot, we found no fixed (bias=3.4 minutes/day; 95%CI=-10.4-17.2) and no proportional (r=-0.2; p=0.09) bias for physical activity (PA); and fixed (bias=-5.3 hour/day, 95%CI=-5.4--5.2; bias=4.5 hour/day; 95%CI=4.3- 4.7) and proportional (r=-0.9; p<0.01; r=0.8; p<0.01) bias for sedentary behaviour (SB) and sleep time, respectively. Using OLP, we found similar findings from baplot for PA (intercept=23.1; 95%CI=-3.04-49.3; slope=0.92; 95%CI=0.83-1.01) and sleep time (intercept=3.14; 95%CI=2.82-3.45; slope=1.20; 95%CI=1.16-1.24). However, we found no fixed and proportional bias (intercept=-0.04; 95%CI=-0.45-0.38; slope=0.20; 95%CI=-0.07-0.10) for SB. Conclusions: OLP could be included in Stata as a valid and comparable alternative to the Bland–Altman method.

and human memory required of common tasks in a complex work environment while speeding up such tasks greatly.

beneficial or that a treatment is safe when it is actually harmful. Propensity-score (PS) matching or weighting adjusts for observed endogeneity, but matching becomes impracticable with multiple treatments, and weighting methods are sensitive to PS model misspecification in applied analyses. We used Monte Carlo simulations (1,000 replications) to examine sensitivity of multivalued treatment inferences to PS weighting or matching strategies. We consider four variants of PS adjustment: inverse probability of treatment weights (IPTW), kernel weights, vector matching, and a new hybrid—vector-based kernel weighting (VBKW). VBKW matches observations with similar PS vectors, assigning greater kernel weights to observations with similar probabilities within a given bandwidth. We varied the degree of PS model misspecification, sample size, number of treatment groups, and sample distribution across treatment groups. Across simulations, VBKW performed equally or better than the other methods in terms of bias and efficiency. VBKW may be less sensitive to PS model misspecification than other methods used to account for endogeneity in multivalued treatment analyses.

identifiers, causing deterministic linking methods (such as Stata's merge or duplicates commands) to fail. For example, one might need to link files that only include inconsistently spelled names, dates of birth with typos or missing data, and addresses that change over time. Probabilistic linkage methods score each potential pair of records on the probability the two records match so that pairs with higher overall scores indicate a better match than pairs with lower scores. Two community-contributed Stata commands for probabilistic linking exist (reclink and reclink2), but they do not scale efficiently. dtalink is a new command that offers streamlined probabilistic linking methods implemented in parallelized Mata code. Significant speed improvements make it practical to implement probabilistic linking methods on large, administrative data files (files with many rows or matching variables), and new features offer more flexible scoring and many-to-many matching techniques. The presentation introduces dtalink, discusses useful tips and tricks, and provides an example of linking Medicaid and birth certificates data.

Markdown tags to Stata’s dyndoc format. This allows the user to write documents in the simple, uncluttered Markdown style used with other programming languages and on websites and still use many of dyndoc’s features such as executing code and embedding graphics links.

every variable in a large dataset can involve a messy set of ad hoc error traps and list commands to learn what unexpected values occur in what dataset rows. Furthermore, code to replace errant values sometimes involves if syntax with a list of terms connected by Boolean ANDs that identify the row targeted for the fix; when typed by hand, these rows are quite susceptible to typographical errors. This talk describes a new command, assertlist, that can test an entire set of assertions in one run without ad hoc code to drill down or move on. Exceptions are listed either to the screen or a spreadsheet. In situations where problematic values will later be corrected or replaced, assertlist generates spreadsheet columns that wait to receive manually corrected values and other columns that immediately put corrected values into Stata replace commands for easy pasting into downstream do-files. In our experience, assertlist streamlines well-documented data cleaning and guards against errors in correction code.

there is a lack of empirical evidence to substantiate a causal relationship. Using novel data on federal regulation of industries (Al-Ubaydli and McLaughlin 2015) and convictions of public officials from the Public Integrity Section, I apply a stochastic frontier approach to account for one-sided measurement error in bureaucratic corruption and the Lewbel (2012) identification strategy to control for potential endogeneity of regulation. Results are striking. Based on the preferred model, there is evidence of endogeneity of regulation and absence of a causal link between regulation and corruption. However, if any of the above two econometric issues are ignored, evidence of a spurious relationship between corruption and regulation is found.

Therefore, it is important to measure patients' satisfaction level with healthcare services in the Emirate to improve the services provided. This study (secondary data analysis) was collected through complex stratified (geographic area), multistage probability sampling. The study examines the satisfaction level with healthcare services in the Emirate of Dubai compared with those who were admitted as inpatients during the last 12 months by using ordered logistics regression. Satisfaction was used as a dependent variable, and many independent variables were used in the model, including suffering from a chronic disease and admission as an inpatient during the last 12 months. Other covariates included were age, gender, insurance type, and nationality. With respect to satisfaction level with healthcare services, we found there is no difference with having or not having a chronic disease, no difference between being male or female, and no difference with the age. All other insurance types are less likely to be satisfied compared with private insurance as a reference group. All other nationalities in Dubai are more likely to be satisfied compared with UAE nationals as a reference group. Not being admitted as an inpatient during the last 12 months in the Emirate of Dubai was more likely to be satisfied with the healthcare services compared with being admitted in the government sector as a reference group. There is need to improve the healthcare services in the Emirate of Dubai in the government sector through public private partnership and competing with the private sector to improve the services among all government health providers, including quality of care and waiting time.

This study measures consumers' premiums for super, good, and normal grades before and after providing grade-labeling information using a nonhypothetical random nth experimental auction. We then estimate consumers' value of grade-labeling information by comparing with market premiums. The results suggest that consumers value the provision of grade-labeling information, with the highest value for the super grade. Given the grade-labeling information, the additional detailed information about grade labeling does not affect consumers' rice-purchasing behaviors. The findings suggest that the rice-grading information is the important factor differentiating domestic rice from imported rice, and it also provides consumers credible information on rice quality to make better purchasing decisions.