Home  /  Users Group meetings  /  2017 Italy

The Italian Stata Users Group meeting was Thursday, 16 November 2017 at the Hotel Brunelleschi, but you can view the program and presentation slides below.

Proceedings

9:30–10:30
Session I: Invited Speaker

Manipulation of SVG graphs with Stata
Abstract: Stata 15 includes a newly documented option to export graphs in scalable vector graphics format (SVG), which is a great storage format because, instead of describing individual pixels, it describes the shapes to be rendered by an application. While other export formats also do this, SVG opens up many more options. Unlike other vector formats used by Stata, SVG is supported by modern web browsers. It is also relatively human-readable, because .svg files are plain text XML code. This opens up a world of opportunity to manipulate SVG files themselves and go beyond the graphics that Stata can currently create.

This presentation gives an introduction to how SVG can be manipulated through several examples: invoking transparent elements 'by hand', hexagonal binning, embedding images to appear behind semitransparent graphs, and including interactive elements in graphs. I hope that this will encourage creative ideas from other users for further extending Stata graphics via SVG.


Additional information:
italy17_Morris.pdf

Tim Morris
University College London
10:50–12:00
Session II: Community-contributed commands and routines, I

Theory and practice of TFP estimation: The control function approach using Stata
Abstract: Alongside instrumental variable (IV) and fixed effects (FE), the control function (CF) approach is the most widely used in production function estimation. Olley–Pakes (OP henceforth), Levinsohn–Petrin (LP), Ackerberg–Caves–Frazer (ACF) have all contributed to the field proposing two-step estimation procedures, while Wooldridge showed how to perform a consistent estimation within a single step GMM framework. In this presentation, we propose a new estimator, based on Wooldridge's, using dynamic panel instruments 'a la Blundell–Bond, and we evaluate its performance by Monte Carlo simulations. We also present a new Stata command prodest for production function estimation and show its main features and key strengths in a comparative analysis with other community-contributed Stata commands. Lastly, we provide evidence of the numerical challenges faced when using OP/LP estimators with ACF correction in empirical applications and document how the GMM estimates vary depending on the optimizer/starting points employed.
Gabriele Rovigatti
University of Chicago Booth School of Business
Vincenzo Mollisi
Free University of Bolzano

ddid: Pre- and post-treatment estimation of the Average Treatment Effect (ATE) with binary time-varying treatment
Abstract: ddid estimates Average Treatment Effects (ATEs) when the treatment is binary and varying over time. Using ddid, the user can estimate the pre- and post- intervention effects by selecting the pre- and post-intervention periods, also by plotting the results in a easy-to-read graphical representation. Also, in order to evaluate the reliability of the causal results achieved by the user's specified model, ddid allows to test both the "common trend" assumption and the degree of "balancing" achieved by the user's specified model. Thus, the model estimated by ddid can be seen as a generalization of the difference-in-differences (DID) approach to the case of many pre- and post-intervention times.

Additional information:
italy17_Ventura.pdf

Marco Ventura
Istituto Nazionale di Statistica

cbps: Stata command to implement the covariate balancing propensity score
Abstract: The dual nature of the propensity score manifests itself in both the conditional probability of treatment assignment and covariate balancing score. The standard approach to propensity score estimates exploits the first feature, leaving the balancing properties to be checked after estimation. Imai and Ratkovic (2014) also focus on the second feature and propose a covariate balancing propensity score (CBPS) estimator that automatically balances the conditional distribution of covariates. Being stated within a GMM framework, CBPS is a simple way to obtain estimates of propensity or weights to be used in subsequent estimations. Monte Carlo studies show its good performance among others in reducing bias of treatment effects estimates. This presentation reviews the method and introduces the Stata community-contributed package cbps, which implements the estimator.
Filip Premik
University of Minnesota
12:00–1:00
Session III: Exploiting the potential of Stata 15, I

Journey to latent class analysis (LCA)
Abstract: Stata's estimation commands have evolved in how they account for groups in the sample. Since the early days of Stata, fitting models with group-specific parameters is simply a matter of using the clause to condition on group membership. In Stata 12, we introduced sem and group analysis for structural equation models (SEMs). Stata 15 introduces two types of group analysis for generalized SEMs. For observed groups, gsem has the new group() option. For latent groups, gsem has the lclass() option and the ability to perform LCA.

Additional information:
italy17_Pitblado.pdf

Jeff Pitblado
StataCorp
2:15–4:00
Session IV: Exploiting the potential of Stata 15, II

Calcolo dell'aderenza alle terapie farmacologiche a partire dai flussi amministrativi correnti
Abstract: Il Medication Possession Ratio (MPR) e la Proportion of Days Covered (PDC) sono le più note misure di aderenza alle terapie farmacologiche derivanti dai flussi amministrativi sanitari. Entrambe esprimono in termini percentuali quanta parte del follow-up individuale del paziente è coperto dal farmaco in studio. L'obiettivo di questo contributo è fornire alcuni consigli per calcolare tali indicatori di farmaco-aderenza usando il software Stata, e presentare alcuni applicazioni pratiche su popolazioni di pazienti affetti da patologia cardiovascolare.

(Presentation to be given in Italian.)


Additional information:
italy17_Lenzi.pdf

Jacopo Lenzi
Alma Mater Studiorum University of Bologna

Matching tra due coorti consecutive estratte da un registro di patologia
Abstract: Utilizzando i dati di un registro di patologia si possono realizzare studi longitudinali che indagano i determinanti di uno o più esiti di interesse per la patologia. In particolare, si può usare una coorte estratta dal registro per eseguire la validazione temporale di un modello predittivo che era stato sviluppato a partire da una coorte di pazienti dello stesso registro arruolati in un periodo precedente. Nella presentazione verrà illustrato il programma Stata che è stato scritto per realizzare un matching a due step, deterministico e probabilistico, con il quale ottenere due coorti omogenee di pazienti in cui confrontare gli esiti di patologia.

(Presentation to be given in Italian.)


Additional information:
italy17_Gibertoni.pdf

Dino Gibertoni
Alma Mater Studiorum University of Bologna

Variabili Socio-Demografiche e Economiche del Voto per Brexit: tendenze territoriali
Abstract: Il 23 giugno 2016 si è tenuto il referendum relativo alla permanenza della Gran Bretagna nell'Unione Europea ("British exit" o "Brexit").

Già nel 1975 i cittadini britannici erano stati chiamati a esprimersi in merito a alla permanenza della Gran Bretagna nell'allora Comunità economica europea. L'affluenza al c.d. "Common market referendum" era stata circa del 65% e il Remain si era affermato in maniera netta (67%) in tutte le circoscrizioni, restituendo l'immagine di una nazione fortemente europeista. I risultati del referendum del giugno 2016 sono stati diametralmente opposti. Anche in questo caso l'affluenza è stata molto elevata (72%), ma a prevalere è stato il Leave con il 52% delle preferenze. Viene fuori l'immagine di un Paese spaccato in due, diviso fra europeisti e antieuropeisti. Cos'è cambiato? Cosa ha spinto i cittadini britannici a esprimersi a favore dell'uscita dall'Unione europea? Quali variabili hanno giocato un ruolo determinante? L'obiettivo di questo lavoro è evidenziare se il voto del 23 giugno sia stato non solo espressione dell'opinione dei cittadini britannici riguardo all'Unione Europea, ma anche e soprattutto la chiara manifestazione di un malessere legato all'influenza di altri fattori, quali la crisi economica e il fenomeno della migrazione, che possono quindi fornire una possibile via per comprendere le ragioni del Leave. In quest'ottica, i risultati elettorali delle local authorities in cui è stato suddiviso il territorio britannico sono stati messi in relazione con una serie di variabili di natura demografica, economica e sociale. Le variabili sono state raccolte o costruite a partire dalle banche dati ufficiali disponibili, in particolare utilizzando le indagini campionarie realizzate da Office for National Statistics, che restituiscono il quadro del Paese al 30 giugno di ogni anno. L'analisi realizzata in questo lavoro fa riferimento alla struttura demografica, economica e sociale del Paese a metà 2015.

(Presentation to be given in Italian.)


Additional information:
italy17_Alaimo.pdf

Salvatore Leonardo Alaimo
Sapienza University of Rome

Socioeconomic factors and the ideal age of union formation. A quantile regression approach
Abstract: There is wide discussion about the relationship between marriage timing and socioeconomic factors. However, empirical evidence on the age at first marriage for women, especially in traditional societies where arranged marriages are common is still limited. We use a nationally representative Qatari Women Survey dataset to test the relationship between ideal age at first marriage and socioeconomic factors. Given the presence of heterogeneity, in preferential age at first marriage for young women, we apply the quantile regression method to make inferences. Using the quantile regression methodology, we find that the influence of socioeconomic factors differs across the ideal age distribution. In particular, there is a positive relationship between the ideal age at first marriage and post-secondary education.
Brian W. Mandikiana
Qatar University
Mahjabeen Ramzan
Qatar University
4:15–5:00
Session V: Community-contributed commands and routines, II

Simulating dynamic panel data in Stata: New features of the xtarsim command
Abstract: The Stata command xtarsim, which I developed in 2005, simulates dynamic panel-data models with exogenous regressors and iid errors. I have now extended the command to simulating models with different types of endogenous or predetermined regressors and with MA (1) errors. This presentation illustrates the new version of xtarsim and presents Monte Carlo applications.
Giovanni S.F. Bruno
Bocconi University

Combining large datasets of patents and trademarks
Abstract: Using the Stata entity names in patents and trademarks have been linked by means of a matching algorithm which accounts for differences due to the position of the same word between otherwise identical strings. In particular, I have taken into consideration the string similarity J index proposed by Thoma, Torrisi, Gambardella, Guellec, Hall and Harhoff (2010), which computes the fraction of common words after breaking up the strings into words at the blank spaces.

Additional information:
italy17_Thoma.pdf

Grid Thoma
University of Camerino
5:00–5:15
Session VI: Report to users & Wishes and grumbles

Workshop: Using simulation studies to evaluate statistical methods

Simulation studies are an invaluable tool for statistical research. They help us understand the properties of statistical methods and compare different methods. To perform a meaningful simulation study, careful thinking needs to be put into planning, coding, analysis, and interpretation. This course will help participants to

  • understand the rationale for simulation;
  • appreciate the importance of careful planning and have a practical framework for planning their own simulation studies;
  • have the tools to code and debug simple simulation studies in Stata;
  • know how to analyze simulation studies producing estimates of uncertainty; and
  • present methods and results for publication.

General information

Examples will be taken from the instructor's experiences in medical statistics. The principles are the same for simulation studies in other applied areas, but the examples may be less relevant.

Participants should be familiar with Stata. For example, they should know how to generate data, run regression commands, and produce simple graphs.

Organizers

Scientific committee

Una-Louise Bell
TStat S.r.l.

Rino Bellocco
Università degli Studi di Milano—Bicocca

Giovanni Capelli
Università degli Studi di Cassino

Marcello Pagano
Harvard University

Maurizio Pisati
Università degli Studi di Milano—Bicocca

Logistics organizer

The logistics organizer for the 2017 Italian Stata Users Group meeting is TStat S.r.l., the distributor of Stata in Italy.

View the proceedings of previous Stata Users Group meetings.