On Monday, I finished looking at the distributions of the interference values using the two different sets of interference values. Last week, I combined the interference values (from the ANOVA) into two sessions in order to have 40 subjects and then divide the 40 subjects into 2 groups based on having interference or not having interference. I completed that step by using a significance test. However, Dr. Bassett wasn't sure that it was the right method, she thought that I might need to use a median split. So, in order to be sure, I emailed Scott and Nicholas a PowerPoint of what I completed and asked what they suggested. They thought using a significance test would be the best option. They especially wanted me to use the interference values from the difference between pre and post movement time because it seemed to follow a normal distribution. Dr. Bassett and I were now certain that the results from Friday were correct.
I then decided to make the community structure I identified two weeks earlier clearer by coloring the nodes on the brain plots more distinct colors, so the difference between the communities was more obvious. I also colored all nodes that were the only node assigned to that community (singletons) grey, so as to get a better understanding of the larger communities. It took me a while to figure out how to threshold the singletons to make them all grey and how to make the communities more distinct colors, but I figured it out in the end. We also had our weekly lab meeting on Monday where a post-doc in the lab, Marcelo spoke about his research.
 |
| This is the brain plot for one scenario (edges from the correlation of interference session 2 and scan 1) with the nodes colored more distinctly. Grey nodes (hard to see) are the singleton nodes. |
Nicholas, one of the researchers I spoke with last Friday sent an email on Tuesday about the plateau results for the motor learning experiment. We decided it would be a good idea to identify whether subjects did or did not have interference when they reached the movement time plateau (no change in movement time greater than 0.25 ms) in order to identify if there was a trend. Attached to the email was a paper, which is not published yet, about the experiment and how the plateau results were found. Nicholas also attached the results from the plateau, including when subjects reached the plateau (which scanning session), which was the most important information in the data he provided. I spent Tuesday reading the paper, looking at the data, and analyzing the results.
On Wednesday, I continued looking at the plateau results and compared those results to whether or not a given subject did or did not have interference. I used the difference between pre and post movement times as the interference values to compare to the plateau results because when looking at the distribution of the interference values, those values gave the most normal distribution. No trend was noticed, there was a mix and the results seemed inconclusive.
The next steps involved looking at the community structure between the two groups in order to see if the structure was different. I completed this step in order to justify that there truly is a difference between the two groups (interference and no interference), it was a necessary step to prove in the paper that the structures are different. Dr. Bassett wanted me to use the interference values from the ANOVA because it more accurately combined the two sequences and found interference values. I worked on going through all the steps: correlation, finding significant edges, permutation test, creating a network, completing 100 optimizations of the community assignments, modular allegiance matrices, and finally consensus partitions for the two groups. I completed these steps the rest of Wednesday and into Thursday.
I had the consensus partitions of the two groups completed by Thursday afternoon, and there seemed to be a difference in the community structures of the two groups (which is a good sign). Dr. Bassett then sent me the movement times of each subject for the two sequences for before each scanning session, during each scanning session, and after each scanning session. The ultimate goal was to analyze the distribution of the interference values for four scenarios, which are described below. I spent the rest of Thursday thinking through how to manipulate all the data. First, I had to find the interference values for the four scenarios for each subject. The four scenarios involved an ANOVA between: pre-scanning session 2 and within scan 2, within scan 2 with post-scanning session 2, pre-scanning session 3 and within scan 3, within scan 3 and post-scanning session 3. I then needed to separate the data for each subject, each sequence so as to make it compatible for the ANOVA. Next, I needed to complete a repeated measures ANOVA for the four scenarios for each subject. This is very tedious, but necessary work. I spent all of Friday working through these steps.
Although I have been here for many weeks, I feel like there is still so much that needs to be done in order to truly have a complete understanding of what causes interference and in order to identify a sub-network that can predict who is susceptible to the interference effect. I hope that within the next two weeks I will be able to accomplish a lot more in moving my project forward. I find this research very exciting and interesting and I am learning so much in the process! I am particularly excited about all the applications my research has like the military, test-taking, and sports. The military is a good example because someone could perform well in practice, but in battle, if that person is susceptible to the interference effect, his/her performance could drop dramatically, which is dangerous for the individual as well as the group. I have developed a deep appreciation for network science and really enjoy the discovery that is involved in research!
No comments:
Post a Comment