SMLG (Statistical Machine Learning Group) Discussion Forum

Class projects correspondences in Advanced Bayesian Inference, Spring 2022

Please see the attachment of my project proposal. I made some changes based on previous commenst.

I calculated the pairwise correlation for all the four datasets.

khasan wrote:Please see the attachment of my project proposal. I made some changes based on previous commenst.

Here are my comments.

Attached is the merged brain tumor data that I will be using for my class project. I merged the 5 individual files that Tanvir cleaned and posted in Dec 2021. The data contains samples for 220 individuals. Originally there were ~60,000 genes in the data files. I was able to eliminate ~14,800 genes that were absent in all 200 individuals.

Here, is the result of 1st two methods. I also complete conditional independence test calculations with no parent, one parent, and two parents. But I am not sure what will be my next step.

cwyoo wrote:

khasan wrote:Please see the attachment of my project proposal. I made some changes based on previous commenst.

Here are my comments.

I try to update the third method. Please take a look and give your valuable feedback.

Here, I attached result for Method Two.

I am having trouble figuring out how to compute BDe scores using SMILE. For now, I'm posting my R code and results comparing BGe scores and BDe scores generated using bnlearn vs. BiDAG for all 25 structures possible for a network that contains 3 randomly selected genes. Also attached is the merged/cleaned brain tumor data that I'm using in my project. When you run the attached R code, you will be prompted to select the file containing the brain tumor data, which you should first save locally on your computer. The code will then output 2 Excel files (saved locally in the directory selected in R). The first Excel file (file name ending in "Scores25Strcts") contains the 4 scores and the differences in the bnlearn and BiDAG scores for each of the 25 structures. The second file contains the ratio of the bnlearn:BiDAG scores, the preferred structure based on the scores, and whether the preferred structure identified using bnlearn and BiDAG agree for all pairwise combinations of the 25 structures.

Here, I attached my updated project proposal. I also attached the results of the three methods.