Dear all, For my research, I need to analyse data consisting of a nominal dependent variable with 38 levels and several independent variables. Most of the IVs are ordinal or nominal with 2 to 5 levels. Some of the IVs are continuous (but I could transform them to categorical variables if needed).
Basically, participants were presented with 12 stimuli (audio or video recordings). For each stimulus, they had to choose a word (in a list of 38 possibilities) that describes the feeling of the person speaking in the stimulus. I am not particularly interested in understanding which participants "got it right" (as in "did perceive the emotion which was intended by the actor"), but which emotion did the participants perceive in each stimulus. For instance, I want to compare the "choice patterns" of the native vs. non-native vs. non-speakers of Chinese, or the "choice patterns" according to the condition in which the stimuli were presented to the participants (and interactions of those variables). Note that I have 1000 participants (so far) and cannot analyse the data via a multinomial regression due to too many empty (or low) cells. A colleague of mine has written a model implementing random forests, which might be a good solution to analyse my data. However, we would like this model to be checked to be sure that it does what it is supposed to do and is valid. I attach the R-script (also available on https://drive.google.com/file/d/1gVypjM3xIz9nxOWpjuDLVUwQIooeDqvV/view?usp=sharing <https://drive.google.com/file/d/1gVypjM3xIz9nxOWpjuDLVUwQIooeDqvV/view?usp=sharing>). The analysis in question starts at line 158 of this script. Many thanks in advance for your help. Kind regards, Pernelle Lorette, PhD Student Department of Applied Linguistics and Communication, Birkbeck University of London p.lore...@mail.bbk.ac.uk __ Take part in my survey on emotion communication! http://bit.ly/EmotionEnglish <http://bit.ly/EmotionEnglish>
