Browsing by Author "Muukkonen, Ilkka"

Multivariate methods make it possible to examine the effects of several variables simultaneously. In cognitive neuroscience, the most frequently used multivariate method is multivariate pattern analysis (MVPA), which has established its place especially in studies using fMRI. MVPA is more versatile and provides better accuracy than the traditional analysis methods. Studies using MVPA can be divided into three categories: studies classifying similar stimuli, studies classifying different stimuli, and representational similarity analysis (RSA). In classification studies the collected data is used to create an algorithm, which is then used to predict observed stimuli. When the observed stimuli are similar to the ones used in creating the algorithm, the accuracy of the predictions can reach remarkably high levels. Using different observed stimuli reduces the prediction accuracy but makes it possible to infer more about the information processing of the brain and improves the ecological validity. Representational similarity analysis allows straight comparison of different stimuli, theoretical models and data from different sources. In RSA, a representational dissimilarity matrix is created from the collected data, and it can be compared to for example the predictions of psychological theories or behavioral results. MVPA-studies have shown that it is possible to get more precise information of the functions of the brain with current imaging methods than was thought to be possible. At their best, multivariate methods can integrate cognitive neuroscience and psychological theories and increase our knowledge of the information processing in the brain.