Psychoco je tradičním setkáním výzkumníků v psychometrice a matematické psychologii. Letošní ročník byl věnován zejména robustním statistickým a psychometrickým metodám pro kategorické a ordinální data. Mezi příspěvky se objevily témata jako bayesiánské síťové modely, různé aplikace a estimátory pro IRT modely, na programu byla i “vášnivá” přednáška o robustní korespondenční analýze. Samozřejmě nechyběly prezentace nových balíčků pro statistický software R, a to např. pro detekci „nepozorného“ odpovídání carelessonset , robustní estimaci mediačních modelů robmed nebo již zmíněné bayesiánské sítě easybgm. A s čím na konferenci přijeli členové INPSY?
Psychoco is a traditional meeting of researchers in psychometrics and mathematical psychology. This year's edition was mainly devoted to robust statistical and psychometric methods for categorical and ordinal data. Papers included topics such as Bayesian network models, various applications and estimators for IRT models, and a "passionate" talk on robust correspondence analysis was on the agenda. Of course, there were also presentations of new packages for statistical software R, e.g., for detecting "inattentive" answering carelessonset, robust estimation of mediation models robmed, or the aforementioned Bayesian network easybgm. And what did INPSY members bring to the conference?
Karel Rečka presented his R shiny application shinyLavaan, which allows easy specification and estimation of confirmatory factor analysis models. The audience was intrigued by the added value of easy model diagnostics through various interactive graphs. Although the software is not publicly available, it will soon be available under an open-source license (in the meantime, interested parties can contact the author). However, the application itself can be tested here.
David Elek presented part of the results of a study investigating the nature of false multidimensionality in psychological questionnaires. The use of the height, weight (Rečka, 2018), and age questionnaire made it possible to reveal the possible nature of the method factors and their non-linear relationship with the construct being measured. The paper itself is currently undergoing reviews and, like the app, should be available to all interested readers in the near future.
Presentation available for downloading
Additional resources
