Conferences

Most of the survey research focuses on individuals, as separate and independent units. The samples are random, seldom include people from the same neighbourhood or particular small social groups. Almost never sampled people know each other, and seldom share the same problems. As J.S. Coleman wrote fifty years ago, which is also true today, individual in many researches is treated as “an atomistic entity” torn out of its social context. This must sound paradoxically in the field of sociology. This problem arises due to usual sample design (random sampling and stratified random sampling) and classical statistical methods that are used in many surveys. Is there any alternative? Large scale educational research (like PISA, TIMS, PIRLS) are some
option here. Those researches use some classical survey methods like questionnaire. Children and parents are asked about classical issues of sociology (social status, attitudes, etc.). The important difference is that whole school or whole classes are sampled. Parents are linked and set in particular context, because of children’s schools. Individuals are not sampled independently, what is considered as an advantage here. Also methods of analysis are somewhat different – multilevel modelling is a main method of analyzing data. Not only relation but variations of relations in different context are taken into account. My aim is to show on some particular examples (from PISA research) how we can see that kind of research as an interesting alternative to
classical surveys based on some kind of simple random sample.