Causal modelling is increasingly being used in both the social and biomedical sciences to model the presence, sign, and direction of influence for the relations of all pairs of variables in a dataset. Typically, causal models are based on structural equations, which are analyzed using regression analysis techniques. The estimated relationships are then, as will be illustrated during the course of the workshop, often mapped in diagrams or flow graphs.
The objective of this workshop is to provide participants with the essential toolset, both applied and theoretical, for the correct implementation of structural equation models (SEM) for statistical causal modelling in Stata. Structural equation models are estimated on both cross-section and time series datasets. The former enabling researchers to undertake both confirmatory factor analysis and causal path analysis. The latter allowing researchers to construct scenario-building and policy simulation and evaluations over time. Although these two methodologies are often treated separately in standard courses, as will become evident, they are in fact intricately linked. During the course of the workshop, participants will also be exposed to the visual intuitive graphical representations of causal links.
At the end of the workshop, participants will be able to autonomously undertake articulated causal designs studies to identify, estimate and test for both direct and indirect causal effects in the presence of unobservable endogeneity, selection bias, measurement error and simultaneity, by implementing either a single-equation causal link (as in an instrumental-variables (IV) setting), or the multi-equation system (SEM) approach. Moreover, participants will take advantage of the availability of time-series data to perform scenario-building and policy evaluation via simulation analysis.
In common with TStat’s workshop philosophy, participants will obtain extensive hands-on experience of the issues under consideration, working on example datasets from both social and biomedical sciences under the careful guidance of the course tutor. Although the course is to be considered primary of an applied nature, technical treatment of the analysis in hand, will however, be provided into order to allow participants to properly address real world applications.
This workshop is of particular interest to biostatisticians, epidemiologists, applied statisticians and researchers or professionals working in the economics, the social sciences or public health.
Basic knowledge of the statistical software Stata.
Knowledge of the following basic statistical concepts: regression model and related properties; point and interval estimation; maximum likelihood estimation.
SESSION I: AN INTRODUCTION TO STATISTICAL CAUSAL MODELLING
Causality in the social and bio-medical sciences: an overview
Marginal, joint and conditional probability
The central role of the conditional expectation
Data generating process (DGP): exogeneity vs endogeneity
Structural, quasi-structural, and reduced-form models
Structural system analysis and its main objects
Structural modelling with cross-section data time-series: Stata sem and forecast packages
SESSION II: AN INTRODUCTION TO THE STRUCTURAL EQUATION MODELLING (SEM) LANGUAGE IN STATA
What is SEM?
Variable definition within SEM
Statistical models using SEM
The sem and gsem Stata commands
The sem syntax
Path syntax using sem
The model _description_options
The option method ( ) and vce ( )
The option covstructure for defining the structure of the variance/covariance matrix
The mathematical notation of SEM
Assumptions under SEM estimation
The Stata SEM Builder
SESSION III: USING SEM FOR CONFIRMATORY FACTOR ANALYSIS (CFA)
What is CFA?
CFA protocol – an illustrative example
Graphical representation of a CFA model
Practical examples using sem for CFA in Stata
SESSION IV: USING SEM FOR CAUSAL PATH-ANALYSIS
Structural equation modelling for path models
Path-model terminology and notation
Exogenous predictor, endogenous outcome, and endogenous mediator variables
Mediation and moderation
Identification and estimation of direct, indirect, and total effects
Recursive and non-recursive models
Estimation of a full structural equation model
Tests for SEM reliability and goodness-of-fit
Revisiting Instrumental-variables (IV) estimation within causal path-analysis
IV identification conditions
Instrument validity and relevance
IV estimation via two-stage least squares (2SLS) using ivregress
SESSION V: SEM PATH-ANALYSIS APPLICATIONS USING STATA
A further look at the implementation of Stata’s SEM packages sem and gsem
The Stata SEM Builder
Using the SEM Builder: a series of illustrative examples
Fitting, modifying and constraining a SEM with sem and gsem
Interpreting the results
Practical examples using Stata
SESSION VI: STRUCTURAL MODELLING WITH TIMES-SERIES DATA USING STATA
Structural modelling with times-series and panel data: an overview
Building time-series structural models in Stata using the forecast package
Model specification and identification
Model estimation: three-stage least squares (3SLS) with the reg3 command
Model validation: static and dynamic forecasts
Practical examples in Stata using real datasets
SESSION VII: STRUCTURAL POLICY EVALUATION
Policy simulation and evaluation via scenario-building using forecast adjust
Dynamic response to exogenous and endogenous shocks
Simulation-based confidence intervals for scenario-building
Practical examples in Stata using real datasets
The course will be held in Berlin, on the 19th, 20th and 21st of September 2018.
Students*: € 735.00
Academic: € 1225.00
Non-Profit/Public Research Centres: € 1513.00
Commercial: € 1800.00
*To be eligible for student prices, participants must provide proof of their full-time student status for the current academic year.
Fees are subject to VAT (applied at the current Italian rate of 22%). Under current EU fiscal regulations, VAT will not however applied to companies, Institutions or Universities providing a valid tax registration number.
Please note that a non-refundable deposit of €100.00 for students and €200.00 for Academic, Non-Profit/Public Research Centres and Commercial participants, is required to secure a place and is payable upon registration. The number of participants is limited to 15. Places will be allocated on a first come, first serve basis.
Course fees cover: teaching materials (handouts, Stata do files and datasets to used during the course), a temporary licence of Stata valid for 30 days from the beginning of the workshop, light lunch and coffee breaks.
To maximize the usefulness of this workshop, we strongly recommend that participants bring their own laptops with them, to enable them to actively participate in the empirical sessions.
Individuals interested in attending this workshop must return their completed registration forms either by email (email@example.com) or by fax (+39 0864 206014) to TStat by the 30th of August 2018.
The objective of this workshop is to provide participants with the essential toolset, both applied and theoretical, for the correct implementation of structural equation models (SEM) for statistical causal modelling in Stata. Structural equation models are estimated on both cross-section and time series datasets. The former enabling researchers to undertake both confirmatory factor analysis and causal path analysis. The latter allowing researchers to construct scenario-building and policy simulation and evaluations over time.