Actuarial Outpost
 
Go Back   Actuarial Outpost > Search Forums
FlashChat Actuarial Discussion Preliminary Exams CAS/SOA Exams Cyberchat Around the World Suggestions


Showing results 1 to 25 of 300
Search took 0.03 seconds.
Search: Posts Made By: NchooseK
Forum: Exam PA: Predictive Analytics 06-15-2019, 03:01 PM
Replies: 584
Views: 76,097
Posted By NchooseK
Took Friday exam. Also used gamma log link...

Took Friday exam.

Also used gamma log link and received a step error on IG, but with another log link. I lazily concluded from the error that gamma(link="log") was my chosen model.

Suffice to...
Forum: Exam PA: Predictive Analytics 06-15-2019, 01:55 PM
Replies: 12
Views: 1,352
Posted By NchooseK
Fitting a GLM or tree to the full dataset is...

Fitting a GLM or tree to the full dataset is optional, but perhaps useful if your data is small or factor levels have few observations. You must be happy with the validation results, first, however.
Forum: Exam PA: Predictive Analytics 06-15-2019, 01:29 PM
Replies: 584
Views: 76,097
Posted By NchooseK
So two more sample exams for the next group. ...

So two more sample exams for the next group. Until SOA changes the exam/syllabus again.
Forum: Exam PA: Predictive Analytics 06-15-2019, 10:47 AM
Replies: 584
Views: 76,097
Posted By NchooseK
Anyone else notice the misordering of factor...

Anyone else notice the misordering of factor levels for at least one predictor in the Data Dictionary?
Forum: Exam PA: Predictive Analytics 06-09-2019, 10:48 AM
Replies: 584
Views: 76,097
Posted By NchooseK
Sorry bud, none of this makes any sense unless...

Sorry bud, none of this makes any sense unless you can supply some context, data, page number or something.

FWIW, you can obtain the coefficient vector of a GLM object in R using either coef(glm)...
Forum: Exam PA: Predictive Analytics 06-09-2019, 01:00 AM
Replies: 10
Views: 770
Posted By NchooseK
That certainly isn't what I want :) But...

That certainly isn't what I want :)

But thank you for the remark.
Forum: Exam PA: Predictive Analytics 06-09-2019, 12:50 AM
Replies: 10
Views: 770
Posted By NchooseK
You make fantastic observations. I like...

You make fantastic observations.

I like the manual idea here b/c SOA is picky, changes its mind, directs us all over creation without a clue what it wants from us, etc.

The modules are...
Forum: Exam PA: Predictive Analytics 06-08-2019, 08:06 PM
Replies: 45
Views: 3,698
Posted By NchooseK
I thought you got your 10 already? Anyway,...

I thought you got your 10 already?

Anyway, you are pretty thorough? What do you have left?

I have glmn, reading partial dependency plots (lol), xgboost and dendro/hierarchical and im DONE. ...
Forum: Exam PA: Predictive Analytics 06-08-2019, 07:12 PM
Replies: 45
Views: 3,698
Posted By NchooseK
Hidden somewhere is a sentence that Tweedie can...

Hidden somewhere is a sentence that Tweedie can be implemented with GLM.

Of course, I haven't seen any other mention.

This means it's 15 points next week!
Forum: Exam PA: Predictive Analytics 06-08-2019, 07:08 PM
Replies: 10
Views: 770
Posted By NchooseK
Goldmine: The Exam PA Manual

Not a pristine 200-page pdf. Not a second job. Not a coming-out party.

A legit PA Manual could make a mint.

It would have be online, able to react to SOA's 11th hour gamechanging and other...
Forum: Exam PA: Predictive Analytics 06-08-2019, 04:32 PM
Replies: 45
Views: 3,698
Posted By NchooseK
Offsets. If total employee hours is an offset,...

Offsets. If total employee hours is an offset, then the business knows in advance how many hours employees are working in a period of time, thus how many hours are at risk per week, month, year,...
Forum: Exam PA: Predictive Analytics 06-08-2019, 04:05 PM
Replies: 45
Views: 3,698
Posted By NchooseK
Offsets wrt to the business problem? Can you...

Offsets wrt to the business problem? Can you give me an example?
Forum: Exam PA: Predictive Analytics 06-08-2019, 03:24 PM
Replies: 45
Views: 3,698
Posted By NchooseK
That was just laziness. Both GLM and DT will be...

That was just laziness. Both GLM and DT will be featured heavily.

SOA actually did a pretty nice job on that sample exam.

The sample solution was outrageous, but easy to follow. Who knows...
Forum: Exam PA: Predictive Analytics 06-08-2019, 02:41 PM
Replies: 584
Views: 76,097
Posted By NchooseK
input variables = predictor variables. ...

input variables = predictor variables.
features = constructed from original independent variables. Features and predictors are similar.
dimensionality = number of levels.
Think 50 US States...
Forum: Exam PA: Predictive Analytics 06-08-2019, 02:35 PM
Replies: 80
Views: 6,457
Posted By NchooseK
"I next tried Race and Gender. I am not...

"I next tried Race and Gender. I am not sufficiently well-versed in the socio-economics of hospital usage, but perhaps there are some interaction effects here. The plot below shows that black males...
Forum: Exam PA: Predictive Analytics 06-08-2019, 01:38 PM
Replies: 584
Views: 76,097
Posted By NchooseK
And then you compare your best models and even if...

And then you compare your best models and even if the PCA wins, you take the other one bc principal components are not in the spirit of interpretability :)

Get your 6 points, say no thanks, move...
Forum: Exam PA: Predictive Analytics 06-08-2019, 12:50 PM
Replies: 584
Views: 76,097
Posted By NchooseK
Piggybacking on the master... PCA is great...

Piggybacking on the master...

PCA is great for few inputs but high dimensionality.
Forum: Exam PA: Predictive Analytics 06-07-2019, 09:08 PM
Replies: 584
Views: 76,097
Posted By NchooseK
"LOS is also a counting variable, but ranges from...

"LOS is also a counting variable, but ranges from 1 to 36. The mean is larger than the median, indicating right skewness. The graph below indicates it [sic] pronounced and thus a log transformation...
Forum: Exam PA: Predictive Analytics 06-07-2019, 07:52 PM
Replies: 3
Views: 270
Posted By NchooseK
Confusion discussed in an old erratum

SOA's official correction on this matter, FWIW.
Forum: Exam PA: Predictive Analytics 06-07-2019, 11:44 AM
Replies: 584
Views: 76,097
Posted By NchooseK
TranceBrah and DjPim should rewrite the modules. ...

TranceBrah and DjPim should rewrite the modules.

At least the parts not copied from Wikipedia.
Forum: Exam PA: Predictive Analytics 06-07-2019, 06:28 AM
Replies: 584
Views: 76,097
Posted By NchooseK
Phuwanut, As someone who HATES learning from...

Phuwanut,

As someone who HATES learning from videos, I found this intro to PCA extremely helpful: https://www.youtube.com/watch?v=_UVHneBUBW0

I then watched several of his other videos.

I...
Forum: Exam PA: Predictive Analytics 06-01-2019, 05:52 PM
Replies: 80
Views: 6,457
Posted By NchooseK
You seem to really command this stuff and I...

You seem to really command this stuff and I appreciate your insights and answers over the past....while.

I agree with your first item. The instructions were explicit that we not over or...
Forum: General 06-01-2019, 04:50 PM
Replies: 2
Views: 1,310
Posted By NchooseK
I kindly ask if ACTEX has fixed, changed--or...

I kindly ask if ACTEX has fixed, changed--or plans on fixing--the digital reader and user interface these products stand on? I haven't been back in a year.

Thanks.
Forum: General 06-01-2019, 04:40 PM
Replies: 7
Views: 563
Posted By NchooseK
Good manuals comprehensively cover all exam...

Good manuals comprehensively cover all exam content as well pesky "prerequisite" topics you naturally may have forgotten. A basic example is of the latter is a worked-out example on matrix...
Forum: Exam PA: Predictive Analytics 06-01-2019, 02:36 PM
Replies: 0
Views: 304
Posted By NchooseK
Collection of Favorite PA Module Slides.

I'll be recreating some of my favorite moments from the PA slides. Up today is Module 6, Slide 54

-----------BEGIN SLIDE 54---------

Interactions

A GLM with only fixed or main effects is...
Showing results 1 to 25 of 300

 

All times are GMT -4. The time now is 01:25 PM.


Powered by vBulletin®
Copyright ©2000 - 2019, 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 0.22975 seconds with 7 queries