Hello All,
I am working on a project wherein we are comparing two methods.
In one method: we see that majority of genes are showing improvement in correlation between predicted and residual expression in comparison to other method.
For example: let says we have 500 genes and are using elastic net and Bayesian method to model gene expression.
method1: 378/500 shows improvement in correlation whereas method2: 122/500: gives good correlation score.
How can we say significantly that method 1 is better in comparison to method2?
Basically, how can I say that 378 genes out of 500 genes their improvement is significant in comparison to method 2?
Is there any way to elucidate this information?
•
written
by
rheab1230 •
0 answers
No answers yet.
Log in to answer this question.
-
snps being shared by gene in different tissues
written by rheab1230 •
-
rank SNPs after elastic net modeling
written by rheab1230 •
-
Hierarchical Clustering using both rows and columns
written by gs000095 •
-
Merge columns adding conditions for every row
written by miriam.riquelmep •
-
Correlation test for multiple variables and adjusted p values
written by ASid •
-
Help with Designing an Approach for Biomarker Signature Generation
written by JJ
-
Troubleshooting assembly contigs of large genome
written by arunprasanna83 •
-
Can or should we perform feature counting with `bedtools intersect -c`?
written by gundalav
-
Elastic net (glmnet) predictive modeling: signal lost in noise
written by lordoftheowl •
-
Tool: Immuno-Navigator, a batch-corrected database for gene coexpression in the immune system
written by alexis.vandenbon •
It's not clear what these methods are doing. Maybe use the mean squared error over all the genes?
Basically both the method produce a gene expression model, we then use the model to predict expression on the remaining datasets and generate a correlation between predicted and actual expression.
Then the MSE would be enough.