Actuarial Outpost
 
Go Back   Actuarial Outpost > Exams - Please Limit Discussion to Exam-Related Topics > SoA/CAS Preliminary Exams > Exam PA: Predictive Analytics
FlashChat Actuarial Discussion Preliminary Exams CAS/SOA Exams Cyberchat Around the World Suggestions


Reply
 
Thread Tools Search this Thread Display Modes
  #91  
Old 11-15-2018, 09:49 AM
DoMathNotMeth's Avatar
DoMathNotMeth DoMathNotMeth is offline
Member
SOA
 
Join Date: Jun 2016
Location: USA
Studying for PA
College: BU Alumnus
Posts: 98
Default

Quote:
Originally Posted by yahoowizard View Post
Is anyone else unable to get into Module 6? I'm able to get into the others, but not Module 6.
I've posted this reply elsewhere in this forum, but I'm repeating it here for your convenience.

I had an issue with Module 6 that prevented me from opening it. I would click "Open Item" then the page would eat up all of my computer's memory before ultimately crashing. Super weird.

Anyways, whether that's your issue or not, I called SOA customer service and they had me send an email describing the issue for their IT department to fix. It took about 48 hours to get my Mod6 access restored so I'd get the ball rolling now.
__________________
Exams: P FM MFE MLC C PA

FAP: 1 2 3 4 5 IA 6 7 8 FA

VEE: Econ Finance Stats
Reply With Quote
  #92  
Old 11-15-2018, 01:13 PM
yahoowizard yahoowizard is offline
Member
 
Join Date: Mar 2011
Posts: 127
Default

Quote:
Originally Posted by DoMathNotMeth View Post
I've posted this reply elsewhere in this forum, but I'm repeating it here for your convenience.

I had an issue with Module 6 that prevented me from opening it. I would click "Open Item" then the page would eat up all of my computer's memory before ultimately crashing. Super weird.

Anyways, whether that's your issue or not, I called SOA customer service and they had me send an email describing the issue for their IT department to fix. It took about 48 hours to get my Mod6 access restored so I'd get the ball rolling now.
I emailed them two days ago, and they responded yesterday by just asking me whether I used a different browser. I called them today and they just told me to email elearn instead. It's annoying since it's the material I'm not super clear on yet and there's not a ton of days till the exam. I hope elearn is quick.
Reply With Quote
  #93  
Old 11-16-2018, 06:41 PM
mistersunnyd mistersunnyd is offline
Member
SOA
 
Join Date: Aug 2016
Studying for how to find a job
Posts: 124
Default

If we have a categorical variable numbered, for example, 0-6, and the model's summary coefficient is like 0.5, do we essentially multiply whatever number in the data (0-6) by 0.5 to get the predicted value? If so, what about the 0? I guess this question also applies to binarization.
Reply With Quote
  #94  
Old 11-16-2018, 09:52 PM
mistersunnyd mistersunnyd is offline
Member
SOA
 
Join Date: Aug 2016
Studying for how to find a job
Posts: 124
Default

Also has anyone tried to manually calculate the predicted face amounts as provided by the Sample Project Report under Findings and Recommendations? So essentially, we multiply $88 by all the factors in the table below, which leads to the predicted face amount? My number was way off when using that formula on my calculator.
Reply With Quote
  #95  
Old 11-16-2018, 10:21 PM
mistersunnyd mistersunnyd is offline
Member
SOA
 
Join Date: Aug 2016
Studying for how to find a job
Posts: 124
Default

Quote:
Originally Posted by mistersunnyd View Post
Also has anyone tried to manually calculate the predicted face amounts as provided by the Sample Project Report under Findings and Recommendations? So essentially, we multiply $88 by all the factors in the table below, which leads to the predicted face amount? My number was way off when using that formula on my calculator.
I think I see what's going wrong. The solution exponentiates each coefficient and asks us to multiply the value from the data by the exponentiated coefficient. What we should be doing for the multiplicative method is exponentiating the product of the coefficient and the data value, such as exp(coefficient for education * number of years of education).

Last edited by mistersunnyd; 11-16-2018 at 10:36 PM..
Reply With Quote
  #96  
Old 11-17-2018, 07:23 PM
NchooseK NchooseK is offline
Member
SOA
 
Join Date: Nov 2012
Location: Philly area
Studying for PA, FAP
College: Swarthmore College (BA Mathematics), Villanova University (MS Applied Stat)
Posts: 242
Default

Quote:
Originally Posted by mistersunnyd View Post
If we have a categorical variable numbered, for example, 0-6, and the model's summary coefficient is like 0.5, do we essentially multiply whatever number in the data (0-6) by 0.5 to get the predicted value? If so, what about the 0? I guess this question also applies to binarization.
If I am interpreting your question correctly, the answer is "no."

A few thoughts:

1) 6 levels is a lot--but if is an important variable, you may have work with it.
2) If the predictor is ordinal, you may want to divide the 6 levels into fewer groups. For example, if the predictor is a survey result on a 0-5 scale and all respondents answering 0 trend one way and the 1-5s trend away from the 0s but similarly to one-another, then you may wish to group 0 and >0 as a flag factor variable. For instance, if the survey asked how frequently do you do meth, fentanyl and heroin at the same time, you can assume someone who has done it once (say, a 1) is pretty messed up like 2s, 3s, etc. Bad goofy example, but perhaps we will all remember it. The 0s are pretty much a completely different animal.
3) Even with ordinal data, you can't always assume that the distance between 0 and 1 is the same as the distance between, say, 1 and 2. Dummy code or manipulate the predictor.
4) The summary coefficient works the way you mentioned if the categorical predictor is binary. So make it binary or don't do that.
5. Unless I am misunderstanding your question, this is something you need to know by now or VERY soon if you want to perform well on the exam.
__________________
Exams: P | FM | C | MFE | LTAM (Confident) | SRM Credit | PA

VEE: Statistics | Finance | Economics

FAP: 1 | 2 | 3 | 4 | IA | 6 | 7 | FA

Conferences: APC
Reply With Quote
  #97  
Old 11-17-2018, 08:47 PM
mistersunnyd mistersunnyd is offline
Member
SOA
 
Join Date: Aug 2016
Studying for how to find a job
Posts: 124
Default

I understand what you're saying, but what if one of the predictors is race and labeled from 0-6? Assuming that race is legal to use as a predictor for some science experiment.

From what I see, you basically just add the coefficient for whatever factor level it is to the formula. Then you apply the log link function.

Last edited by mistersunnyd; 11-17-2018 at 10:19 PM..
Reply With Quote
  #98  
Old 11-17-2018, 10:55 PM
Totient(43) Totient(43) is offline
Member
SOA
 
Join Date: Apr 2017
Posts: 54
Default

Quote:
I understand what you're saying, but what if one of the predictors is race and labeled from 0-6? Assuming that race is legal to use as a predictor for some science experiment.
If you have a categorical variable and you want to do some regression you can use dummy variables. For high, medium, and low risk, you use n-1=2 dummy variables. One possible way to define them is
x_1 x_2
high 1 0
med 0 1
low 0 0



So you have E(y)=b_0+b_1x+b_2x_2 and from our definitions we have:
E(y|high)=b_0+b_1(1)
E(y|med)=b_0+b_2(1)
E(y|low)=b_0

If you had something like y=.2+.5x_1+.4x_2 and you want to make a prediction for someone in the high-risk group, E(y|high)=.2+.5.
__________________
P FM MFE C something else
Reply With Quote
  #99  
Old 11-19-2018, 11:21 PM
Demo99 Demo99 is offline
Member
SOA
 
Join Date: Jan 2013
Location: CA
Posts: 145
Default

Quote:
Originally Posted by mistersunnyd View Post
If we have a categorical variable numbered, for example, 0-6, and the model's summary coefficient is like 0.5, do we essentially multiply whatever number in the data (0-6) by 0.5 to get the predicted value? If so, what about the 0? I guess this question also applies to binarization.
If you have a categorical variable, shouldn't you be using them as factors and creating 6 dummy variables?
Reply With Quote
  #100  
Old Yesterday, 09:43 AM
DarchArvor's Avatar
DarchArvor DarchArvor is offline
Member
SOA
 
Join Date: Jun 2016
Studying for everything
Posts: 480
Default

Quote:
Originally Posted by mistersunnyd View Post
I understand what you're saying, but what if one of the predictors is race and labeled from 0-6? Assuming that race is legal to use as a predictor for some science experiment.

From what I see, you basically just add the coefficient for whatever factor level it is to the formula. Then you apply the log link function.
Say you have colors (and not races):

Set Variable color = C

Then C is:

Blue = 0
Yellow = 1
Green = 2
Pink = 3

Then you have to create 3 dummy variables:

C_1 = 1 if blue, 0 otherwise
C_2 = 1 if yellow, 0 otherwise
C_3 = 1 if green, 0 otherwise

You don't have to create C_4, because if all C_1, C_2 and C_3 equal 0, then it automatically means you predicted that the color is pink.
__________________
P FM MFE C MLC PA FAP
Reply With Quote
Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off


All times are GMT -4. The time now is 01:39 PM.


Powered by vBulletin®
Copyright ©2000 - 2018, Jelsoft Enterprises Ltd.
*PLEASE NOTE: Posts are not checked for accuracy, and do not
represent the views of the Actuarial Outpost or its sponsors.
Page generated in 0.30789 seconds with 9 queries