I conducted a thirty day experiment with about 10 participants. The experiment was supposed to test whether one can "train" humans to predict the age(a), height(h), and weight(w) of a person from their voice.
I conducted two experiments: 1) Regression where I asked participants to assign a numerical number to each of the 5 speaker's a,h,w.; 2) Binary classification where I asked participants to assign who is taller, heavier, and older from two pairs of speakers.
What kind of interesting analysis can be done from this data?
Interesting project.
For 1), you could use the Mean-squared error averaged across the particpants, for each time point and look for a trend. This trend can be statistically estimated using ANOVA to test if your results are time-dependent or not.
For 2), you could do the same but this time with the accuracy value.
Could you tell us if the 10 participants belong to different groups that could be used to comapre them?
Thanks Kevin.
Within the thirty day experiment, there were two sets of experiments as such 1) For around 10-15 days, we revealed the answers of all the participants to all the participants as a group at the end of the experiment once they completed the entire set of experiments, and 2) We revealed the answers of each question immediately after answered. So there was a difference in the time of 'feedback'.
Other than that, the participants were not really explicitly assigned to different groups, but they did belong to different nationalities, different age groups -- which could perhaps be used to "quantify" perception.
If you want to test whether the different feedback strategies have an impact, you can follow what I said earlier with a statistical test at each time point.
- Badges
- Science topic