|09:50-10:00||Introduction and welcome from the scientific organisers
|10:00-10:40||Modeling unobserved heterogeneity in Stata
Rafał Raciborski, StataCorp
Unobserved heterogeneity refers to differences among individuals that cannot be measured by regressors. In this presentation, I will talk about models where unobserved heterogeneity is assumed to follow a discrete distribution.
Such models are called finite mixture models or latent class models. Stata 15 introduced the
fmm prefix for fitting finite mixture models. More complicated models for latent classes can be fit using
gsem with the
lclass() option. This is a nontechnical talk with an emphasis on concepts and no prior knowledge of the presented material is assumed.
|10:40-11:00||New Stata procedures
University of Warsaw
Analyzing large-scale achievement surveys in Stata using PISATOOLS and PIAACTOOLS
Large-scale achievement surveys like PISA, PIAAC, PIRLS or TIMSS use complex sampling and scaling methods. While these methods allow reliable comparisons of student or adult achievement across countries, they also pose a barrier in terms of data analysis especially for researchers less experienced in programming. Although Stata allows usage of replicate weights and takes into account complex survey design for most commands, the use of plausible values is less straightforward. In my presentation, I will discuss how to properly estimate basic statistics and regression models with PISA data. I will focus on common mistakes made by researchers who try to find shortcuts in their analyses of large-scale achievement surveys. I will then present two packages who I co-authored that facilitate analysis of the commonly used surveys: PISA and PIAAC.
The recently updated PISATOOLS package contains several commands that facilitate analysis of the data from the OECD PISA study. These commands allow analysis with plausible values and derive standard errors using the BRR method implemented in PISA. The command
pisastats allows calculating basic statistics like mean, median, percentiles, standard deviation etc. The commands
pisacmd allows using several regression and estimation commands. The commands
pisaoaxaca facilitate decomposition analysis with the PISA data. All commands save output files as HTML tables that can be easily opened in spreadsheet programs or internet browsers but also save results in Stata matrices. The PIAACTOOLS package helps analyzing PIAAC data that follows similar methodology as PISA. At the end of my presentation, I will also discuss other packages that facilitate analysis with similar data and will give example of short programs that help using these datasets with Stata estimation commands that do not follow typical syntax accepted by our package.
|11:00-11:20||Uncertainity and sensistivity analysis of composite indicators using Stata
Composite indicators (CI) are increasingly used in many fields, in particular, in policy evaluation and public communication (eg. HDI, AAI, etc.). They are used to benchmark countries performance in fields such as economic activity, population well-being, technological development or ecology.
CI are usually constructed as a weighted linear combination of relevant normalised one-dimensional sub-indicators. There is a great deal of uncertainty in the process of constructing every CI, as always an alternative set of sub-indicators, alternative weighting system and alternative method of aggregations could have been adopted. In the presented paper a new Stata procedure is presented which enables researchers to assess the level of uncertainty and sensitivity of results to changing assumptions in the process of constructing CI.
|11:50-12:10||Modelling the link between energy security and international competitiveness
Honorata Nyga-Łukaszewska & Eliza Chilimoniuk-Przeździecka
In the world of open economies and free trade, countries are strongly focused on gaining and maintaining the ability to compete with their products successfully in the international market. The objective of this paper is to identify how energy security affects trade competitiveness. It is a crucial point for understanding of the energy security phenomenon. It enables us to verify whether energy security is just a goal on itself or it can be a factor determining economic performance more broadly than only GDP. Initially, energy security research were solely focused on macroeconomic activity depicted by GDP performance (e.g. Leiby, Jones & Curlee, 1997).
Using Stata framework, we investigate the link between energy security and trade and assess energy security effect on trade competitiveness in defined groups of countries. This relationship is assessed in the manufactured goods grouped according to the BEC classification (Broad Economic Categories), which presents end-use categories, which are meaningful within the framework of the System of National Accounts (SNA). The study includes 23 countries denoted by one of the world’s biggest energy consumption levels between 1995 and -2014. In our research we use data from the Institute for 21st Century Energy provided by the U.S. Chamber of Commerce, World Bank, and OECD.
|12:10-12:30||(Mis)use of matching techniques
University of Warsaw
Matching techniques became very popular among researchers in recent years due to ready to use commands embodied in statistical packages. However, they are not “magic bullet” that solves all statistical problems. The idea of matching is simple: modify Your data in such way that it can be treated as a result of completely randomized experiment. Several matching methods and algorithms are proposed and discussed in the literature. The problem is that practitioners either are not aware or ignore their shortcomings.
During the presentation most popular matching methods will be discussed, their statistical properties and important limitations. Numerical examples will be provided.
|12:30-12:50||Weighting Sub-Populations in Longevity Inequality Research: A Practical Approach
Warsaw School of Economics
The weights allowing calculation of life expectancy for a whole population as a weighted average of group-specific life expectancies are proposed. They are characterized by a minimum distance from the actual population shares that are different from those assumed in life tables. It is demonstrated how they may be obtained by means of constrained regression, using Stata. The problem of negative solutions is also addressed. The empirical examples include longevity inequality calculations under various weighting systems. The data come from the Human Mortality Database and from Russia’s regional statistics.
|13:00- 13:30||Wishes & Grumbles Session & Final Remarks