Menu

2017 Stata Autumn School, London

  • Location: Cass Business School, Bunhill Row, London EC1Y 8TZ
  • Duration: 6 days (2nd October 2017 - 7th October 2017)
  • Software: Stata
  • Level: Advanced, Intermediate
  • Delivered By: Dr Karla Diaz-Ordaz (LSHTM); Dr Luigi Palla (LSHTM); Dr Ruth Keogh (LSHTM); Prof Aurelio Tobías (IDAEA-CSIC); Dr Matteo Quartagno (LSHTM)
  • Topic: Autumn School, Medical statistics, Meta-analysis, Statistics, Survival analysis, Various methods
2017 Stata Autumn School, London

Course Overview

Our Stata Autumn School comprises a series of six, 1-day courses delivered by experienced biostatisticians: Dr Ruth Keogh, Dr Karla Diaz-Ordaz, Dr Luigi Palla and Dr Matteo Quartagno all of the London School of Hygiene and Tropical Medicine (LSHTM) and Prof Aurelio Tobias from IDAEA-CSIC (Barcelona). Each course will include an initial introduction to the topic followed by hands-on examples. There will be plenty of time dedicated to interpretation of the results, discussion of assumptions and comparison of approaches.

This is a great opportunity for students, academics and professionals to expand their statistical skills and learn how they can apply statistics from biostatisticians at the forefront of their specialist fields. The combination of separate courses allows complete flexibility to register only the courses that they find most relevant to their research interests.

The separate courses comprising the Stata Autumn School are:

  • Course 1: Dealing with missing data
  • Course 2: Analysis of survey data
  • Course 3: Survival analysis
  • Course 4: Analysis of hierarchical data
  • Course 5: Structural equation modelling
  • Course 6: Network meta-analysis

Timberlake Consultants are the official Stata distributors to the UK, Ireland, Middle East and North Africa, Spain, Portugal, Poland and Brazil.

Course Agenda

Course 1: Dealing with missing data

Date: Monday, 2 October 2017
Delivered by: Dr Karla Diaz-Ordaz, LSHTM
Learning ratio: 50% theory and 50% practical

We begin by illustrating with a simple dataset the adverse consequences missing data can have on inferences. Next we give an intuitive explanation of Rubin’s classification scheme for missingness mechanisms (MCAR, MAR, MNAR), and explore how missingness mechanisms can be described and investigated using Stata. We then move on to a brief discussion of the deficiencies in several commonly used ad-hoc approaches to handling missing data before we introduce the method of multiple imputation (MI), a principled approach for handling missing data under the MAR assumption. Both joint model and chained equations imputation will be described, and we will apply these to data using Stata's MI commands. We briefly introduce an alternative approach to handling missing data, that of inverse probability weighting, and illustrate how this is readily performed in Stata, and conclude by emphasising the important role of sensitivity analyses when analysing partially observed datasets.

  • Session 1: Impacts of missing data, classifying missingness mechanisms and ad-hoc methods
  • Session 2: An introduction to multiple imputation (MI) (single variable)
  • Session 3: Multiple imputation for multiple variables (chained equations)
  • Session 4: Inverse probability weighting and sensitivity analyses after MI

Course 2: Analysis of survey data

Date: Tuesday, 3 October 2017
Delivered by: Dr Luigi Palla, LSHTM
Learning ratio: 45% theory and 55% practical

We will describe the general statistical techniques that apply to almost all forms of survey data. We will start by introducing the most common sampling designs used in collecting survey data, such as simple random, cluster and stratified sampling designs. Their main features, such as sampling weights, clustering and stratification will be reviewed. The course will cover one, two and multiple stage survey designs, and the three variance estimators implemented in Stata's survey estimation commands. We will introduce the svyset command which declares the data to be complex survey data, specifies the variables that identify the survey design characteristics, and the default method for variance estimation. The course will also cover the estimation approaches using the svy: prefix, which are implemented in the Stata survey family of commands. Lastly we will cover post-stratification, a method for adjusting sampling weights to account for underrepresented groups in the population, and the analysis of strata with one sampling unit and those with certainty units. Examples and exercises will use datasets from the Stata Survey Documentation.

  • Session 1: Sampling design characteristics.
  • Session 2: Special types of sampling units.
  • Session 3: Poststratification and Regression with survey data
  • Session 4: Variance estimation

Course 3: Survival analysis

Date: Wednesday, 4 October 2017
Delivered by: Dr Ruth Keogh, LSHTM
Learning ratio: 45% theory and 55% practical

We will discuss various approaches to the analysis of time-to-event data. These are data that arise from following up individuals until a particular event is observed, or until their follow-up is interrupted (i.e. time is censored). Hence, in this setting, the outcome of interest consists of two pieces of information: the time which the subject spends in the study and what happens at the end of this time. We will first introduce the powerful stset command which declares the key survival variables. We will then discuss censoring mechanisms, and the estimation and comparison of survival curves using the Kaplan-Meier, the life-table (actuarial) method, and the log-rank test. The course will also cover estimation of the cumulative hazard function using the Nelson-Aalen estimator. The two most commonly used multivariable models for survival analysis - Cox and Poisson models - will be introduced and compared using real data examples taken from clinical and epidemiological studies.

  • Session 1: Introduction to survival analysis
  • Session 2: Kaplan-Meier survival curves and the log-rank test
  • Session 3: The Cox proportional hazards model
  • Session 4: Checking the proportional hazards assumption

Course 4: Analysis of hierarchical data

Date: Thursday, 5 October 2017
Delivered by: Dr Matteo Quartagno, LSHTM
Learning ratio: 50% theory and 50% practical

We will introduce the course with examples of settings where the usual assumption of independent units of analysis does not hold. If this dependency is ignored, any subsequent inferences are potentially invalid. Dependency therefore must be dealt with. We will discuss alternative approaches to achieve this, with the focus on methods that explicitly specify the nature of the dependency, i.e. mixed effects models. The alternative generalised estimating equations approach will also be briefly described. The focus will be on continuous outcomes when there are only two levels of aggregation. As mixed effects models are a development of ANOVA and the linear regression model this is where we will start. We will then introduce random intercept models, with and without covariates, using thextmixed command, and then more general random coefficient models. Throughout we will discuss assumptions and ways to assess the appropriateness of the fitted model. An example of modelling individual measurement repeated over time (longitudinal data in biostatistics) will conclude the day.

  • Session 1: Impact of dependency, choice of strategies and revision of linear regression
  • Session 2: The random intercept model
  • Session 3: The random coefficient model
  • Session 4: Models for longitudinal data

Course 5: Structural equation modelling

Date: Friday, 6 October 2017
Delivered by: Dr Matteo Quartagno, LSHTM
Learning ratio: 50% theory and 50% practical

In this course we will give a brief introduction to generalisations of standard regression models that deal with multivariate outcomes, i.e. structural equation models (SEM). These models include path analysis and factor analysis. We will start by introducing path analysis, i.e. the joint analysis of an outcome and any of its predictors which we also wish to model. The predictors that are also modelled are denoted "intermediate" outcomes, while the main outcome of interest is termed "distal". We will discuss and fit path analytical models using the command sem on data taken from life course epidemiology. We will also introduce the concepts of direct and indirect effects as used in the SEM literature, highlighting their limitations. Factor analysis will be introduced as an approach to deal with measurement error in a variable of interest when there are error-prone observations available. Such models can again be fitted using sem. Examples will be used, and assumptions and generalizations discussed. More general SEMs where a path model includes latent variables will conclude the day.

  • Session 1: From univariate to multivariate models: notation and diagrams
  • Session 2: Path regression
  • Session 3: Confirmatory factor analysis
  • Session 4: Linear structural equation modelling

Course 6: Indirect Comparisons and Network Meta-analysis with Stata

Date: Saturday, 7 October 2017
Delivered by: Prof. Aurelio Tobias, Spanish Scientific Research Council
Learning ratio: 30% theory, 20% demonstration and 50% practical

A one-day course that is aimed at both academics and practitioners, with a basic knowledge of Stata, who are interested in applying network meta-analysis using Stata commands designed for this purpose.

Course Overview:

Meta-analysis has traditionally been used to synthesize the effectiveness of an intervention from a collection of studies. However, when there are no studies directly comparing two or more interventions, traditional meta-analysis cannot estimate their comparative benefits. Although if there is information available regarding the effectiveness of two interventions, named B and C, in comparison to a common comparator A, an indirect treatment comparison may be used to estimate a comparison of the effectiveness of B compared with C. Approaches to meta-analysis have been increasingly implemented to estimate the effects of multiple interventions, taking into account the full network of available studies and simultaneously incorporating direct and indirect comparisons.

This one-day course introduces the main statistical techniques to analyse a network meta-analysis in practice using the network suite of Stata commands. It is aimed at both academics and practitioners, with a basic knowledge of Stata, who are interested in applying indirect comparisons and network meta-analysis using Stata commands designed for this purpose.

The course will cover the following:

  • Types of comparisons in meta-analysis
  • Indirect comparisons with meta-regression
  • Multivariate meta-analysis
  • Full network meta-analysis
  • Practical exercises using the network suite of commands

Pre-course readings

  • Course 1:
    • Sterne, J., et al., (2009). Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls, 338: b2393, BMJ.
    • Schafer, J. L., (1999). Multiple imputation: a primer, 8: 3-15, SMMR.
  • Course 2:
    • StataCorp, (2013). Data-Management Reference Manual, Stata Press.
  • Course 3:
    • StataCorp, (2013). Survival Analysis and Epidemiological Tables, Stata Press.
    • Kleinbaum, D., and Klein, M., (2011). Survival Analysis, a Self-Learning Text, Springer.
  • Course 4:
  • Course 5:
    • StataCorp, (2013). Structural Equation Modeling, Stata Press. (Stata 13 or Stata 12 release).
    • Kleinbaum, D., and Klein, M., (2011). Survival Analysis, a Self-Learning Text, Springer.

Suggested Readings

  • Course 1:
    • Van Buuren, S., (2007). Multiple imputation of discrete and continuous data by fully conditional specification, 16: 219-242, SMMR.
    • Carpenter, J. R., & Kenward, M. G., (2012). Multiple imputation and its application, Wiley.
  • Course 2:
    • Levy, P., and Lemeshow, S., (1999). Sampling of Populations, Wiley.
  • Course 3:
    • Cleves, M., Gould, W., Gutierrez, R. G., and Marchenkov, Y. V., (2010). An Introduction to Survival Analysis Using Stata, 3rd Ed., Stata Press.
  • Course 4:
    • Rabe-Hesketh, S., and Skrondal, A., (2012). Multilevel and Longitudinal Modeling Using Stata, Volume I: Continuous responses, 3rd Ed.. Stata Press.
  • Course 5:
    • Introductory: Kline, R. B., (2004). Principles and Practice of Structural Equation Modeling, 2nd Ed., New York: Guildford.

    • Advanced: Skrondal, A., and Rabe-Hesketh, S., (2004). Generalized Latent Variable Modeling, Boca Raton, Fla: Chapman.

Course 1: Dealing with missing data

Participants should have a working knowledge of Stata, and in particular be familiar with regression models, such as linear and logistic regression, and their interpretation. No knowledge regarding missing data techniques will be assumed.

Course 2: Analysis of survey data

Knowledge of Stata is not required, but attendees are assumed to have some statistical knowledge, such as what is typically covered in an introductory statistics course. Participants should be computer literate, able to manage files and familiar with Microsoft Windows.

Course 3: Survival analysis

Knowledge of Stata is not required, but attendees are assumed to have some statistical knowledge, such as what is typically covered in an introductory statistics course. Participants should be computer literate, able to manage files and familiar with Microsoft Windows or Mac OSX.

Course 4: Analysis of hierarchical data

Basic knowledge of Stata required and familiarity with linear regression models and the regresscommand in Stata.

Course 5: Structural equation modelling

Basic knowledge of Stata required and familiarity with linear regression models and the regresscommand in Stata.

  • Student registrations: Attendees must provide proof of full time student status at the time of booking to qualify for student registration rate (valid student ID card or authorised letter of enrolment).
  • Additional discounts are available for multiple registrations.
  • Cost includes course materials, lunch and refreshments.
  • Delegates are provided with temporary licences for the software(s) used in the course and will be instructed to download and install the software prior to the start of the course. (Alternatively, laptops can be hired for a fee of £10.00 (ex. VAT) per day).
  • If you need assistance in locating hotel accommodation, please notify us at the time of booking. Please note: do not book accommodation or travel until the course has been confirmed
  • Payment of course fees required prior to the course start date.
  • Registration closes 5-calendar days prior to the start of the course.
    • 100% fee returned for cancellations made over 28-calendar days prior to start of the course.
    • 50% fee returned for cancellations made 14-calendar days prior to the start of the course.
    • No fee returned for cancellations made less than 14-calendar days prior to the start of the course.

The number of delegates is restricted. Please register early to guarantee your place.

  •  CommercialAcademicStudent
    1-day pass (02/10/2017 - 07/10/2017)
    2-day pass (02/10/2017 - 07/10/2017)
    3-day pass (02/10/2017 - 07/10/2017)
    4-day pass (02/10/2017 - 07/10/2017)
    5-day pass (02/10/2017 - 07/10/2017)
    6-day pass (02/10/2017 - 07/10/2017)

All prices exclude VAT or local taxes where applicable.

* Required Fields

£0
- +
Post your comment

Timberlake Consultants