Page 3 of 3

Re: Probabilistic Graphical Models, Fall 2016

PostPosted: Tue Dec 06, 2016 12:31 pm
by lsand039
Here is my simplified structure:
Log (p)= -4790.529381
Confidence:
min 0.01113009
max 0.105499243

Validation results:
Test:
Alzheimer = 0.741325 (235/317)
Healthy = 0.705128 (110/156)
Diseased = 0.776398 (125/161)
ROC AUC: Healthy: 0.798216
ROC AUC: Alzheimer: 0.798236

Leave One Out:
Alzheimer = 0.697161 (221/317)
Healthy = 0.762821 (119/156)
Diseased= 0.63354 (102/161)
ROC AUC: Healthy: 0.742017
ROC AUC: Alzheimer: 0.742017

4-fold Cross Validation:
Alzheimer = 0.678233 (215/317)
Healthy= 0.775641 (121/156)
Diseased= 0.583851 (94/161)
ROC AUC: Healthy: 0.746835
ROC AUC: Alzheimer: 0.746715

Re: Probabilistic Graphical Models, Fall 2016

PostPosted: Wed Dec 07, 2016 9:30 am
by lsand039
Here are the results from the best structure:
Log(p)= -4781.533175

Validation results:
Test:
Alzheimer = 0.741325 (235/317)
Healthy = 0.705128 (110/156)
Diseased = 0.776398 (125/161)
ROC AUC: Healthy: 0.798216
ROC AUC: Alzheimer: 0.798256

Leave One Out:
Alzheimer = 0.697161 (221/317)
Healthy = 0.762821 (119/156)
Diseased = 0.63354 (102/161)
ROC AUC: Healthy: 0.742017
ROC AUC: Alzheimer: 0.742017

4-fold Cross Validation:
Alzheimer =0.690852 (219/317)
Healthy= 0.782051 (122/156)
Diseased =0.602484 (97/161)
ROC AUC: Healthy: 0.732541
ROC AUC: Alzheimer: 0.732541

Re: Probabilistic Graphical Models, Fall 2016

PostPosted: Thu Dec 08, 2016 12:25 am
by igrif005
Good evening,

Please see my attached presentation for class tomorrow morning.

Isabel

Re: Probabilistic Graphical Models, Fall 2016

PostPosted: Thu Dec 08, 2016 1:42 am
by igrif005
Evening,

Please see my attached Final Paper. Looking forward to discussing our projects in class tomorrow.

Isabel

Re: Probabilistic Graphical Models, Fall 2016

PostPosted: Thu Dec 08, 2016 10:35 am
by igrif005
GeNIe Model.

Re: Probabilistic Graphical Models, Fall 2016

PostPosted: Thu Dec 08, 2016 11:52 am
by igrif005
Updated HIV status node to reflect prevalence in Florida and sexual orientation node to reflection national distribution (at this point in time I haven't found any Florida-specific data).

Re: Probabilistic Graphical Models, Fall 2016

PostPosted: Fri Dec 09, 2016 9:27 pm
by igrif005
Hi Dr. Yoo,

Here is the final draft of my paper incorporating the changes from the other day.

Re: Probabilistic Graphical Models, Fall 2016

PostPosted: Thu Dec 15, 2016 11:07 am
by cwyoo
igrif005 wrote:Hi Dr. Yoo,

Here is the final draft of my paper incorporating the changes from the other day.


Good job. For publication, I will suggest you should describe more about Bayesian networks, , e.g., relationships in serial, merging, and diverging connections, etc.