University of Twente Student Theses

Login

Small and negative correlations among clustered observations: A simulation study about limitations of the linear mixed effects model

Nielsen, N.M. (2019) Small and negative correlations among clustered observations: A simulation study about limitations of the linear mixed effects model.

[img]
Preview
PDF
501kB
Abstract:Usually, the linear mixed effects model is used for analysing clustered data. The model assumes that observations in clusters are positively correlated. When there is a true negative correlation within groups, the model assumes incorrectly that observations are independently distributed. A simulation study was conducted to report about the biases that occur when ignoring negative correlation within groups. The results show that, when ignoring the negative correlation, the Type-I error is deflated, and standard errors and p-values are overestimated. This resulted in an underestimation of the difference between group means and the population mean. These findings indicate that researchers need to be aware that negative correlation between observations in clustered data may occur. Also, it is important to watch out for biases when ignoring negative correlation within groups and incorrectly assuming that observations are independently distributed. Altogether, this study shows that the linear mixed effects model is an inappropriate tool for analysing clustered data containing negative correlation within groups. Keywords: linear mixed effects model, negative correlation, negative ICC, Type-I error
Item Type:Essay (Bachelor)
Faculty:BMS: Behavioural, Management and Social Sciences
Subject:70 social sciences in general, 77 psychology
Programme:Psychology BSc (56604)
Link to this item:http://purl.utwente.nl/essays/78254
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page