»  Home »  Stata Conferences »  2021 Mexico

Proceedings

10:00–11:00 Introduction to Bayesian VAR estimation in Stata Abstract: The use of the Bayesian approach for regression analysis is spreading more across different disciplines.
... (Read more)
The possibility to incorporate a priori information in the form of probability distributions for the parameters of the model makes this approach highly appealing when the researcher has that knowledge. Bayesian vector autoregressive models (BVAR) are particularly attractive because the overparameterization present in many VAR models can be handled by using prior probability distributions that allow shrinking of the parameter space. In this presentation, I will briefly highlight the general elements associated with Bayesian VAR models, and I will use an example for a macroeconomic VAR model to illustrate the way Stata implements the estimation for a VAR model using the Bayesian approach.

(Read less)

Additional information:
Mexico21_Sánchez.pdf

Gustavo Sánchez
StataCorp
11:00–11:20 Maps and matrices on population and housing census indicators Abstract: In this presentation, we discuss various exercises that include calculus and comparison of indicators, graphic facilities, estimation of Mincerian econometric models, and
... (Read more)
basic procedures to report matrices and maps on economic indicators, sociodemographic indicators, and profiles about the characteristics of the population and housing at the Mexican regions, states, and municipalities, using Stata and Mata commands, from the available original microdata samples of the 2010 and 2020 Censuses of Population and Housing conducted by the National Institute of Statistics and Geography (INEGI). The syntax, matrix results, and templates that are presented show the versatility of Stata as an ideal tool in the management and analysis of large volumes of data with a focus on assessment issues, statistics, and econometrics that require the use of real and recent information to be analyzed or summarized with algorithms, procedures, and structured code.

(Read less)

Additional information:
Mexico21_Islas.pptx

Juan Francisco Islas
IPN
11:20–11:40 Data management techniques in long and wide formats and analysis of food patterns by principal components Abstract: The food frequency of consumption questionnaire (FFQ) is a semiquantitative instrument with a structure that allows the transference from long to wide format employing factorial analysis by principal components to identify patterns of healthy eating by stage of formation in a population of undergraduate students of nutrition at the Universidad Autónoma del Estado de Morelos (UAEM).
... (Read more)
A cross-analytical study carried out between 2017 and 2018 with 563 students applied the FFQ validated for ENSANUT 2016, capturing the information in long format. A database with equivalent quantitative data was conformed to calculate the total of calories consumed. The principal variables were identified and converted to wide format to analyze using Stata 14. An exploratory factor analysis was made to assess the students’ feeding patterns. We identified differences by stage of formation. The food patterns of undergraduate students of nutrition at UAEM change from the basic to the discipline stage. Stata is an ideal program for this type of analysis, and the management of data is fundamental.

Contributors:
María Alejandra
Ortega Ceballos
Paola Adanari
Zúñiga Hernández
Ofmara Yadira
Morales Vences
Yesenia Sánchez Rabadán
Ana Gabriela
Universidad Autónoma del Estado de Morelos
(Read less)

Additional information:
Mexico21_Terrazas-Meraz.pdf

María Alejandra Terrazas-Meraz
Universidad Autónoma del Estado de Morelos
11:40–12:00 Analysis of the socioeconomic factors affecting the prevalence of diabetes in Mexico using Stata Abstract: In recent years, chronic and degenerative diseases in Mexico such as diabetes have shown an exponential increase, mainly in populations vulnerable by level of income.
... (Read more)
Mexico ranks sixth worldwide and the first in Latin America on obesity prevalence. Mexico has high social inequality, with about 50% of the population with some level of poverty, situations that contribute to the development of diabetes. We employ the National Health and Nutrition Survey (ENSANUT) to investigate the socioeconomic variables, such as household living conditions and education level, influencing the prevalence of diabetes in the Mexican population by urban or rural strata, the geographical position, and some health comorbilities. Using multivariate statistics and multiple correspondence, we built an wellness index, later employed to develop an integrating analysis of logistic regression to model the presence or absence of diabetes diagnoses as a function of the socioeconomic covariables and control variables. The study concludes that a better material well-being and higher level of education are factors of protection in the face of diabetes development in the Mexican population.

Contributor:
Carla Gómez Fernández
Universidad Católica de Chile
(Read less)

Additional information:
Mexico21_Martínez-Bautista.pdf

Humberto Martínez-Bautista
CIMAT
12:20–12:40 Mexican teenagers: Violence in couple relations Abstract: Objective: Describe the prevelance, type of violence, and factors associated with the violence in couple relations (VP) of teenagers in Mexico.
... (Read more)
Methodology: Cross-section of the National Consumption of Drugs, Tobacco and Alcohol 2016 survey. A sample of 3,862 individuals between 12 and 19 years old explores the distribution of several variables employing logistic regression to estimate odd ratios (OR). The analysis was made using Stata 15.

Results: 52.63% of the population are female, and the mean VP in teenagers was 11.49%, with the greatest prevalence in the State of Mexico with 16.17%. The greatest prevalence of VP on females was detected in Sinaloa (14.65%) and on men in the state of Nuevo León (15.26%). The prevalence of psychological, physical, and sexual violence was of 11.12%, 1.52%, and 0.51%, respectively. The factors associated with greater VP in teenagers are to be women (OR=1.82, [CI95% 1.13-2.93]); absence of schooling (OR=1.80 CI95% [1.06-3.06]); living in a metropolitan area (OR=1.65 CI95% [1.05-2.59]); length of relationship (OR=3.32 [CI95% 1.44-7.65]); and overconsumption of alcohol (OR=4.17 [CI95% 2.35-7.39]).

Conclusions: The prevalence of violent behavior in the relationships of couples among teenagers is an urgent problem that needs to be studied at different levels.

Contributors:
Terrazas-Meraz
María Alejandra
Macías-Carrillo Claudia
Universidad Autónoma del Estado de Morelos
(Read less)

Additional information:
Mexico21_Adanari.pptx

Ortega-Ceballos Paola Adanari
Universidad Autónoma del Estado de Morelos
1:00–1:20 An application to tobacco tax reforms using sids and duvm Stata commands Abstract: This presentation illustrates an application of a relatively new set of Stata commands based on two specifications of the Deaton models: The Single Item Deaton System (sids) and the Deaton Unit Value Model (duvm).
... (Read more)
The application estimates in the first part, the own and cross-price elasticities of price–demand for tobacco, alcohol, and sugary drinks in the general population and for subpopulations of consumer groups. Subsequently, vulnerable groups in health in Mexico are observed when matching two surveys (Enigh and Ensanut), specifically those suffering from obesity, diabetes, and hypertension and obtained for tobacco consumers by using the duvm command to obtain their own rate elasticities. Data used come from the Encuesta Nacional de Ingresos y Gastos de los Hogares (ENIGH) during 2018, with deflated current prices and incomes for 2020. Both commands grant simplicity at implementation, and results from tobacco tax reforms in Mexico show that the three products are recently complementary because they have a significantly negative elasticity, and their prices increasing cause consumption reduction rates to increase the most within the diabetic group.

Contributors:
Abdelkrim Araar
Université Laval
Linda Llamas
Universidad Estatal de Sonora
(Read less)

Additional information:
Mexico21_Huesca.pdf

Luis Huesca
CIAD
1:40–2:40
Keynote presentation
Estimation after stepwise in Stata: Why and how? Abstract: This talk discusses some Stata commands that produce valid inference for causal parameters after selecting which covariates should be included in the model.
... (Read more)
The talk also provides a quick introduction to using the lasso and to using the BIC-based stepwise methods of covariate selection and some tradeoffs between them. It also discusses some of the methodology implemented in the discussed Stata commands, including some new results in Drukker and Liu (2021).

Contributor:
Di Liu
StataCorp
(Read less)

Additional information:
Mexico21_Drukker1.pdf

David Drukker
Sam Houston State University
10:00–11:00 Customizable tables Abstract: Presenting results effectively is a crucial step in statistical analyses, and creating tables is an important part of this step.
... (Read more)
In this presentation, I will discuss the use of two new features introduced in Stata 17: The updated table command and the new collect suite. Traditionally, the table command was used to compute and display summary statistics. The updated version not only allows us to do that but also allows us to customize those tables and create new tables with regression results, among others. In addition, collect allows us to create customized tables incorporating returned elements from any Stata command. To create a table, you can start with a simple layout that specifies only the contents in columns and rows and then use a variety of features to customize the look of your table, controlling elements such as labels, text alignment, border lines, numerical format, and much more. This can be done either from the command line or using the new Tables Builder, which gives you direct access to the different features. After creating and customizing a table, you can export it to Word®, Excel®, LaTeX, PDF, Markdown, HTML, SMCL, and plain text. I will describe the main elements you need to know to start creating your own tables, and I'll guide you through some examples to illustrate the process.

(Read less)

Additional information:
Mexico21_Canette.pdf

Isabel Canette
StataCorp
11:00–11:20 Dealing with endogenous treatment conflated with sample selection and measurement error Abstract: We consider the case of the estimation of a (causal) treatment effect in a continuous response model with sample selection and where the binary treatment is endogenous and subject to measurement error.
... (Read more)
This econometric model, or "structure", is common to many applications of interest in economics. However, today many applied economists are unsure what must be done to obtain a consistent estimator and give up. In this presentation, we review in detail the consistent estimators available in this context and discuss their properties in finite samples using a Monte Carlo study, such as the one reported in the online appendix by Miranda and Yu (2020).

(Read less)

Additional information:
Mexico21_Miranda.zip

Alfonso Miranda
Centro de Investigación y Docencia Económicas (CIDE)
11:20–11:40 Using Stata and Python for nonparametric smoothing estimation Abstract: The choice of the binwidth/bandwidth (smoothing parameter) is one of the most relevant issues of density estimation.
... (Read more)
For univariate data, cross-validation (CV) is a well-known procedure for automatic choosing the smoothing parameter. Least-squares CV minimizes an estimation of the integrated squared error by means of the “leave-one-out estimate”, a computationally intensive method ideal for WARPing implementation. On the other hand, by directly estimating the asymptotic mean integrated squared error, it obtains a biased estimation called biased cross-validation (BCV). With asymmetrical or heavy-tailed distribution, BCV tends to oversmooth while L2CV, despite having larger dispersion, on average, produces better estimations. In regression, there are several estimators of discrepancy between the estimation and the true curve. In this respect, from several methods, the leave-one-out cross-validation stands out. In this talk, we present ado-file–Python script combinations to calculate univariate kernel density estimators and kernel regression (Nadaraya–Watson). We also include programs to calculate L2VC and BCV for univariate bandwidth choosing and a program that employs a combination of several penalizing functions and cross-validation algorithms to select bandwidth in kernel regression based on the algorithms from Scott (1992; 2015), Härdle (1991), Salgado-Ugarte (1995; 2002) and Salgado-Ugarte & Saito-Quezada (2020).

Contributor:
V. M. Saito-Quezada
M. M. Salgado-Saito
N.I. Plascencia-Díaz
FES Zaragoza UNAM
(Read less)

Additional information:
Mexico21_Salgado-Ugarte.zip

Isaías Hazarmabeth Salgado-Ugarte
FES Zaragoza UNAM
11:40–12:00 Presidential approval, democracy, and accountability in Mexico (1994-2018) Abstract: Although a bunch of studies suggest that the perception of corruption undermines citizens’ attitudes toward political institutions and regime legitimacy, less attention has been given to this: How important is perception of corruption for presidential approval?
... (Read more)
More precisely, how different are the effects of citizens’ perceptions of the president´s competence to deal with corruption on approval before, during, and after a democratic transition? By using Mexico as a case study and using a Wald test with the Stata command, this presentation shows that after a critical election that leads a democratic transition, the president’s competence to deal with corruption has stronger effects on approval for Big Mammals (the new elite) than for Dinosaurs (presidents of the old authoritarian party) and this stronger effect on approval survives even when the authoritarian party returns to power. Democratic processes' expectations explain this: With a democratic transition, citizens evaluate the forthcoming presidents by both their competence to manage the economy and their competence to combat corruption.

(Read less)

Additional information:
Mexico21_Gómez-Vilchis.ppt

Ricardo Román Gómez-Vilchis
UAM Cuajimalpa
12:20–12:40 getmxdata: Importing data from INEGI and Banxico direct to Stata Abstract: getmxdata is a Stata command that allows one to import data from the National Institute for Statistics and Geography (INEGI), specifically the time-series data available in the Banco de Inromación Económica (BIE) site and the Sistema de Información Económica (SIE) of Banco de México (the Central Bank), directly into Stata.
... (Read more)
Using the Application Programing Interface (API) of these two institutions, getmxdata allows the user to import the selected variables, assessing a brief description and obtaining a variable in the date format and with the respective frequency. These web services of public information are increasingly common. The Federal Reserve System in the United States and the World Bank are well-known instances. There are commands to import data from these sources but no software that allows importing information directly from INEGI or Banxico to Stata until now.

(Read less)

Additional information:
Mexico21_González_Favila.zip

Miguel Angel González Favila
Universidad Autónoma de Ciudad Juárez
12:40–1:00 ras2dta, spwmatrix, spmat, and xsmle commands to approximate economic growth using satellite images in Mexico Abstract: I present a measure to approximate economic growth based on the methodological proposal by Henderson et al. (2012, AER), with data generated through satellite images instrumenting economic variables.
... (Read more)
The command is particularly useful because of the lack of timely information on national accounts at the local level in Mexico. Using proxy variables can complement this lack of information and help to measure the economic dynamics at subnational levels through changes in observed luminosity from outer space. We document a positive relation between economic growth and predicted growth from luminosity. Using these commands, we generate ASCII-raster grids produced in ArcView to feed the proxy indicator of economic growth through nightlights captured by satellite imaging. We simultaneously apply a two-stage panel-data spatial Durbin model linking GEODA extensions to account for the time space nature of economic dynamics.

(Read less)

Additional information:
Mexico21_Millán_López.pdf

Andrés Jerson Millán López
Universidad de Guadalajara
1:00–1:20 An alternative for variable selection in high-dimensional linear regression models Abstract: Chudik, Kapetanios, and Pesaran (Econometrica 2018. 86: 1479-1512) propose one covariate at a time, multiple testing (OCMT) for variable selection in linear regression models of high dimension as an alternative approach to penalized regression of the lasso type.
... (Read more)
In this presentation, we replicate key OCMT results using Stata instead of the original programs in MATLAB. Using the new commands in Stata developed for baing and ocmt, we find that our results are consistent with the Monte Carlo simulations by Chudik, Kapetanios, and Pesaran. In addition, we exactly replicate their findings in the empirical application. We also find the same five variables with the highest frequency of inclusion based on the OCMT selection method.

Jesús Otero
CIDE
(Read less)

Additional information:
Mexico21_Núñez.pdf

Héctor M. Núñez
CIDE
1:40–2:40 A simple solution to the simultaneous inference problem Abstract: Stata commands that implement frequentist estimators produce output tables that report multiple hypothesis tests and multiple confidence intervals.
... (Read more)
Each of these tests and each of these confidence intervals is individually reliable, but simultaneously interpreting more than one of them is misleading. This talk explains the issues and presents a postestimation command that produces hypothesis tests and confidence intervals that can be interpreted simultaneously.

(Read less)

Additional information:
Mexico21_Drukker2.pdf

David M. Drukker
Sam Houston State University
2:40–3:00 Open panel discussion with Stata developers
StataCorp

Keynote presentation

Title

Estimation after stepwise in Stata: Why and how?

Presenter

David M. Drukker

Description

This talk presents a few commands in Stata that produce reliable inferences for causal parameters after choosing which covariates should be in the model. This talk also provides a basic introduction to the lasso method and the BIC-stepwise method that can be used to choose which covariates should be in the model. I compare the costs and benefits of the two methods. I also present a bit of the statistical methodology that is implemented in these commands, including some new results in Drukker and Liu (2021).

Visit the official conference page for more information.

Scientific committee

Alfonso Mendoza Velázquez (Chair)
CIIE & UPAEP
Willy W. Cortez Yactayo
CUCEA & Universidad de Guadalajara
Alejandro López Feldman
CIDE
Alfredo Cuecuecha Mendoza
CIIE & UPAEP

Logistics organizer

The logistics organizer for the 2021 Mexican Stata Conference is MultiON Consulting S.A. de C.V., the distributor of Stata in Mexico, Latin America, and the Caribbean.

View the proceedings of previous Stata Conferences and Users Group meetings.


Stata

Shop

Support

Company


The Stata Blog: Not Elsewhere Classified Find us on Facebook Follow us on Twitter LinkedIn YouTube Instagram
© Copyright 1996–2021 StataCorp LLC   •   Terms of use   •   Privacy   •   Contact us
× 20% off Gift Shop purchases! Sale ends 12/11 at 11:59 PM CT. Use code GIFT20.