Actuarial Outpost June 2019 Exam PA
 Register Blogs Wiki FAQ Calendar Search Today's Posts Mark Forums Read
 FlashChat Actuarial Discussion Preliminary Exams CAS/SOA Exams Cyberchat Around the World Suggestions

 DW Simpson International Actuarial Jobs Canada  Asia  Australia  Bermuda  Latin America  Europe

#251
05-29-2019, 09:30 AM
 Gettin Lucky In Kentucky Member SOA Join Date: Apr 2018 Location: Louisville Ky Studying for Specialty College: Eastern Kentucky University Graduate Posts: 32

Is anyone planning on taking an FSA exam in the Fall? Will you begin studying right after this exam is taken? I have everything on my transcript for ASA except PA and APC (July) and don't really want to waste time.
#252
05-29-2019, 11:05 AM
 TranceBrah Member SOA Join Date: Mar 2014 Location: Best Coast Posts: 240

Quote:
 Originally Posted by Gettin Lucky In Kentucky Is anyone planning on taking an FSA exam in the Fall? Will you begin studying right after this exam is taken? I have everything on my transcript for ASA except PA and APC (July) and don't really want to waste time.
I'll be doing LPM in Oct. The good thing about PA timing is that you don't need to know your result before deciding to start studying for next exam. If you fail, then you have all of November to study for PA again.
#253
05-29-2019, 12:13 PM
 Josh Peck Member SOA Join Date: Dec 2016 College: Towson University Posts: 99

Quote:
 Originally Posted by noone Thanks. So the examples in module 7 use issage (predictor) for the x-axis and yhat on the y-axis. What does y-hat mean? The target variable is claim count and is either C (actual_cnt=0) or N(actual_cnt>=1).
The y axis is the log odds.
This is because, for GLMs the linear part equals g(mean).
Where g() is the link function.

Note that When the log odds = 0, the probability = 0.5
__________________
P FM MFE C PA
#254
05-29-2019, 12:15 PM
 Josh Peck Member SOA Join Date: Dec 2016 College: Towson University Posts: 99

Quote:
 Originally Posted by Josh Peck It makes no difference to use sensitivity vs specificity as long as you know how to interpret it, which it seems that you do. It simply depends on which factor level is labeled as TRUE.
Does anyone know how to easily switch the levels though?
While it doesn't really matter, it is an inconvenience.
__________________
P FM MFE C PA
#255
05-29-2019, 02:45 PM
 Adapt and Chill Member SOA AAA Join Date: Sep 2017 College: Davidson College Posts: 189

Quote:
 Originally Posted by Josh Peck Does anyone know how to easily switch the levels though? While it doesn't really matter, it is an inconvenience.
Yeah, you can pass an argument for the positive class. Per ?confusionMatrix:

Quote:
 positive - an optional character string for the factor level that corresponds to a "positive" result (if that makes sense for your data). If there are only two factor levels, the first level will be used as the "positive" result.
## Default S3 method:
confusionMatrix(data, reference, positive = NULL,
dnn = c("Prediction", "Reference"), prevalence = NULL,
mode = "sens_spec", ...)
#256
05-29-2019, 03:06 PM
 Josh Peck Member SOA Join Date: Dec 2016 College: Towson University Posts: 99

Quote:
 Originally Posted by Adapt and Chill Yeah, you can pass an argument for the positive class. Per ?confusionMatrix
Nice! I never took note of the argument before
__________________
P FM MFE C PA
#257
05-29-2019, 03:11 PM
 Josh Peck Member SOA Join Date: Dec 2016 College: Towson University Posts: 99

Does anyone have info on how to pick a good cutoff value for classification predictions?

I would assume you use the ROC curve and pick a cutoff such that your cutoff is close to 0 without seeing a dropoff in the ROC Curve.
However I have no idea how to actually use this idea in practice.

Edit:
I actually found an answer to this fairly quickly.
https://stats.stackexchange.com/ques...imple-function
__________________
P FM MFE C PA

Last edited by Josh Peck; 05-29-2019 at 03:58 PM..
#258
05-29-2019, 04:31 PM
 Yossarian Member SOA Join Date: Jun 2011 Location: SoCal Posts: 120

Quote:
 Originally Posted by Adapt and Chill Yeah, you can pass an argument for the positive class. Per ?confusionMatrix: ## Default S3 method: confusionMatrix(data, reference, positive = NULL, dnn = c("Prediction", "Reference"), prevalence = NULL, mode = "sens_spec", ...)
another way, using the "Student Success" solution code for the variable Train.DS\$G3.Pass.Flag:

> contrasts(Train.DS\$G3.Pass.Flag)
.. P
F 0
P 1

the contrast table shows you that Fail = 0, Pass = 1, with binarization of F left out

you can relevel this with :

> contrasts(relevel(Train.DS\$G3.Pass.Flag,"P"))
.. F
P 0
F 1

now Fail = 1, Pass = 0, with binarization of P left out

I THINK. I believe contrasts shows you how R "dummifies" the levels ...

Last edited by Yossarian; 05-29-2019 at 04:52 PM.. Reason: Added .. to the P and F, as they weren't aligning above their respective columns
#259
05-29-2019, 09:34 PM
 Gettin Lucky In Kentucky Member SOA Join Date: Apr 2018 Location: Louisville Ky Studying for Specialty College: Eastern Kentucky University Graduate Posts: 32

Quote:
 Originally Posted by TranceBrah I'll be doing LPM in Oct. The good thing about PA timing is that you don't need to know your result before deciding to start studying for next exam. If you fail, then you have all of November to study for PA again.
Those were my thoughts as well. Thanks for the confirmation. 2019, Thanks for handing me another year without having a life.
#260
05-30-2019, 05:51 PM
 noone Member SOA Join Date: Feb 2017 Posts: 138

Does anyone know if we are able to use the ? function that exists withing R during the exam?