Conclusion¶
This study involved analysing the data from the Iowa Gambling task across 10 different studies with 617 participants. Initial data analysis showed a clear trend of choices away from the bad decks and towards the good decks which rewarded contestants who chose consistently across all trials. There was some interesting participants choices such as two participants who selected the same deck for 150/150 trials but the vast majority showed an ability to learn throughout the study and choose the good decks more often than not.
Our clustering analysis focused on the clusters across payoff schemes (k=3) and across studies (k=10). The three payoff schemes were all different in some aspects of their rewards and punishments. After analysing the data we found that payoff scheme was the largest contributing factor to the clusters followed by the study within which our participants came from.
Future work¶
This assignment was a very simple approach to clustering with no prior practical knowledge of k-means. Hence this analysis is quite basic in nature but provided a good insight into k-means and its potential uses and limitations. In the future it would be good to further explore this data set and address the health of individuals and how this affects their choices over time.