Johan Lyhagen

Professor of Statistics. Conduct research mostly in the area of time series econometrics. I work, for example to develop models and tests for unit roots and cointegration. Do also research in psychometrics dealing with issues of non-normality and ordinal variables.

From a statistical point of view, economic equilibriums are often modeled as linear relations. These are called cointegrated relationships. There are many examples of economic equilibriums; a person consumes on average almost all their income, the price level of a country is about the same as its neighbor if the price level is adjusted for the exchange rate. Basic questions are how to empirically examine the existence of equilibriums between economic variables, if there are, how can they be estimated? There is a significant effort to gather data on region and country level. Part of my research concerns how data from multiple regions/countries can be used in cointegration analysis.

I'm also researching in psychometrics, especially what is called structural equation models. Many psychological phenomena are not directly observable, they are latent. The classic example is intelligence measured by IQ tests. Another common example is socioeconomic status within sociology that can be measured by education, income and occupation. Usually, the observable variables do no fulfill the conditions that are normally assumed. They can be ordinal (i.e., A is greater than B, but we do not know how much more) or skewed. This will in various ways to influence the properties of the statistical methods used. How methods are affected and what we can do about it are parts of my research area.

My entry in the Mathematics Genealogy Project can be found here.