Psychiatry - Theses

Permanent URI for this collection

http://hdl.handle.net/11343/233

Filters

2018 1
Reset filters

Settings
Statistics
Citations
Type: PhD thesis × Start date: 2018 × End date: 2018 × Author: de Araujo, Luis Fernando Silva Castro ×

Search Results

Now showing 1 - 1 of 1
  • Item
    Thumbnail Image
    Dimensional constructs and methods for their treatment in psychiatry research
    de Araujo, Luis Fernando Silva Castro ( 2018)
    Psychiatry has lagged behind other fields within the cognitive sciences regarding the analysis of dimensional constructs. This has surely improved recently by diagnostic classifications’ organisation of communication around mental disorders, however it has not kept up with developments in psychology in recent decades. Statistical methods developed for the study of intelligence could have been incorporated in psychiatry, as they may improve the field by permitting objective evaluation of dimensional constructs. For several reasons, this path has not been taken. This project aims to ameliorate this issue, by trying to do just that. It first applies a familiar method in psychology called structural equation modelling (SEM) to specify latent variables in data sets with information from subjects suffering from mental disorders. Genetic and neuroimaging (fractional anisotropy and volumetric segmentation) data is used and incorporated into structural equation models. Secondly, the project analyses what this approach can offer psychiatry, and what limitations of these techniques one should expect. It was found that these methods may help in specifying relations between psychometric scores and neurological features that are not reductive, or at least are not radically so. These methods are more parsimonious when comparing data from neuroimaging and data from cognitive abilities. When variables representing cognitive scores are involved in multiple comparisons, they may overestimate significant correlations, because of the shared variability that these types of cognitive variables present. SEM helps to improve that. Furthermore, factor analysis may be a good method of defining latent constructs that should fit appropriately with new proposals for classification like the recent Research Domain Criteria. Finally, psychiatry may benefit from a picture of science that accepts modelling as a viable approach for describing and understanding psychiatric conditions.