SAS Certification Base/Advanced/Predictive-Modeling: Study Strategy (About Statistical Software)

SAS certifications Strategy (Programming vs. Predictive)

Check out the Link: http://qcfinance.in/sas-business-analytic-course/

I heard about this software by some people on LinkedIn, and also saw its requirements on various job portals. The exam seems simple and doable, now will look into more details. We have 2 approaches one is learning database commands and other is the real thing which is prediction like that on decision trees.

Introduction: I have been reading about SAS and found it one of the most powerful tools which include power of C++ and MATLAB (linking to things I know). I will be discussing about SAS base certificate, studies, strategies, syllabus and how much time I will be allocated to study this area. The first part is about base programmer which is just like database management.

SAS Base (easy and data management): In this regard I had a look at the book "SAS-Base-Certification-Preparation" which is a 500 page big book and found it quite useful, I will try to join a coaching where I can do hands on practice on SAS and also try to implement my older models on research in SAS. The courses that I saw are of around 50 hours for Base, and hence I think I need around 100 hours in total to prepare for this exam. The 7 chapters in the book does not include any financial or complex models, but rather are very very basic and describes the overall framework, hence there is no fin in the initial stages to give this exam.

There are 2 levels in this course for software engineering:

1. Base Exam is about introduction (OOC, managing data, generating reports, managing input output, acquaintances to platform etc).
2. Advanced Exam which is to be given after the base exam includes Macros and SQL integration.

How to start preparation, and general matters. All programming languages has these things:

1. OOC remains the same.
2. Data type follows general ways.
3. Input-output.
4. Referencing.
5. Library.
6. Database extraction.

Coming to more important part i.e. Predictive Modeling Using SAS. This is a tough but a very important which has decision tree and regression.  This is the tool for finance people and risk managers. 

Broad areas at stat packages that you need to work independently on SAS not covered in certification:

1. ANOVA
2. Regression
3. Time series analysis.

Ref: http://en.wikipedia.org/wiki/Comparison_of_statistical_packages

Current strategy:

I have downloaded some presentations and trying to look for books to start with. I am having moderate exposure to stat, but know all the stat that will be used in Pred Modeling. I plan to give the exam in Jan. last week. I plan to join some coaching where I can be taught by SAS Pred Model certified teachers and practice on SAS installed at his coaching.

Broad areas that I need to master in Stat Analysis, if possible using SAS:

Linear regression, logistic regression, survival analysis, neural network, time series analysis, conjoint analysis, clustering techniques, decision trees and linear programming etc.

Financial Modeling  on other platforms (Excel, Macros, VBA, SAS, MATLAB):

I was looking at 2 options on financial modeling: SAS and Excel. I will taught by the current insti over the next 2 months on Excel, so I am working to get the SAS base certificate. It seems to be a 180 USD exam, and will explore more on that. SAS is mentioned at Berk, NYu, Qnet as useful so I think moving into it will be worth.

I also read about SAS CFA interlinking for a CFO on the CFA website: SAS for the CFO: Helping CFOs Adjust to an Expanding Role. And it did make a lot of sense because CFO should be able to link the performance matrix.

SAS Predictive Modeling contents and Strategy:

1. Accessing and Assaying Prepared Data (typical data management activity).
2. Decision Trees (here it is your own decision tree).
3. Regressions (you know this).
4. Neural Networks and Other Modeling Tools (highly researched topic).
5. Model Assessment and Implementation (requires tool).
6. Pattern Discovery (theoretical topic).
7. Memory based reasoning (this is something new, I will add notes on this).

The thing here is that you can still understand many of things without even giving your hand on Predictive modeling. The things here that are important are mostly derived from statistics. Neural network was a subject in engineering especially for CSE people, and it was a topic of 8th sem of RGTU people. I focus on how this thing is used in financial engineering. The tool that most people use in this regard is the MATLAB Toolbox.

Conclusion of the visits to over 4 coaching and official training provider of SAS in Jan 12:

• SAS can be learned quickly if you know the maths and stat.
• SAS base and advanced are more of programming type, SAS predictive is specific to engineering.
• SAS predictive modeling can be learned quickly if you know the major 4 areas that are also a part of the engineering syllabus.
• SAS predictive modeling course is 30 hours by these institutes, and teachers are scarce to find, and for these 30 hours they charge huge money. This money is 2.5 times my PGDF fee, 2.5 times my CFA coaching fee, 2.5 times of what I earn (all stat in per hour), thus it is an expensive program by all means.
• The part which can give trouble is macros, SQL which is used in Predictive modeling as well.
• For resume purpose, I think you can still give Base SAS which is programming, but does it help really in the Quant job that one needs to get?

MATLAB Finance toolbox:

I have some experience about this software when I did reliability engineering, and also in electrical modeling. Now I will check out the MATLAB Fin toolbox and see if it works on not.

Conclusion: Language not tough for those acquainted with MATLAB and C++, and can be learned quickly. Base has no great thing and has to cover both levels. SAS predictive is the one that makes the most sense in Financial Engineering, but SAS base talks about basics of data handling. In the long term SAS, predictive modeling will help, but for now even SAS base certificate will be good enough to prove that you are acquainted with the platform.

Versions at friends and coaching institutes:

I went to check but I found SAS Data miner for Predictive modeling no where, it is in fact the best software I want to learn today. But to learn Base and Advanced SAS is very easy, also because it is something related to database management and nothing related to forecasting.

Here is a brief summary of the software tools that I liked to work on in Finance:

1. For derivative pricing: MATLAB and Mathematica.
2. For Stat: SAS and R (There are certifications in SAS but R is open source and no certi exists as per my findings).
3. For common day to day use: VB and Excel.

Other tools for predictive modeling (only):

Enterprise miner, Knowledge seeker, Treenet.

Free Trial :http://www.salford-systems.com/downloadspm.html.

Target Date: Jan 12 Last week or July 12; Fee 180 USD for base, 250 for Predictive Modeling, similarities with OOC and MATLAB.

R vs. SAS Dilemma and my views:

1. About R Software: This software is open source and is having a big community to do research. There is no certification in this exam, otherwise this package is very good as it is open source.
2. The problem with R is that unlike SAS it has no certification course but has got  a huge support online. SAS is more seen as requirement than R and SAS seems to be more user friendly. Hence things are bit tricky, personally I like open source and things which has got a lot of help and effort online. But here I will prefer SAS due to job requirements and proof of knowledge as SAS certifications. 
3. I still don't know the answer of whether predictive modeling can be done in R, in the same way it is done in SAS. I think there are ways to do all things in R but the paths can be longer.

Will try to compare it with Oracle Hyperion Financial Management.

What training has to offer and how to do that on your own?

Training is important part, but SAS training is very expensive and you can still learn it or go for open source options like R. CFA Case is about data management, I had a look and it was all playing with data, no maths involved. Advanced looked into Macros and SQL, and this is again hardcore programming. For these 2 things, you can do on your own as you can find a place to practice yourself.

Total training time = 3 months full time for SAS B+A+P which means around 300 hours.

120 for Base, 120 for Adv, 80 for Predictive. Looking at this you can also make a chart and do it on your own.

Another parallel strategy is to do it on R.

Sample STAT-Quantitative Job requirements:

-Statistical expertise and knowledge of few of the following techniques – linear regression, logistic regression,   survival analysis, neural network, time series analysis, conjoint analysis, clustering techniques, decision trees   and linear programming etc.,

- Expertise in the usage of SAS for both data manipulation and model development.

- Experience with modeling software – Eg: Enterprise miner, Knowledge seeker, Treenet etc.

Some areas of Artificial Intelligence that are used:

• Fuzzy logic
• Artificial Intelligence
• Genetic Algorithm
• Neural Network
• Monte-Carlo Method
• ACO (Ant Colony Optimization)
• AHP(Analytic Hierarchy Process).

Some Derivatives pricing Models:

• Rational pricing assumptions
• Moneyness
• Pricing models
• SABR Volatility Model
• Markov Switching Multifractal
• The Greeks
• Finite difference methods for option pricing
• Trinomial tree
• Optimal stopping (Pricing of American options)
• Interest rate derivatives
• Short rate model
• Hull–White model
• Cox–Ingersoll–Ross model
• Chen model
• LIBOR Market Model
• Heath–Jarrow–Morton framework.

Risk Analytics has SAS Analytics Statistics includes but not limited to:

• PD, LGD models with hands on experience in creating models.
• Basel framework and regulations and experience in creating related models. 
• Running SAS queries to prepare datasets used in analysis and predictive modeling. 
• Using SQL from SAS to extract and aggregate data from larger data sources. 
• Performing ad-hoc analysis/statistical analysis and generation actionable reports. 
• Idea about SAS/SPSS/MS Excel/MS Access and VBA etc. 
• Major tools include analysis tools like SPSS, R.
• IT data management tools and BI platforms.

Links:

http://www.puzha.com/sasbook/sas%20examples.html

http://www.ats.ucla.edu/stat/sas/webbooks/reg/chapter1/sasreg1.html.

SAS Base Course (making it relevant to Financial Risk)

Uploaded by Shivgan on WizIQ Tutorials

One on One Customized Training:

qcfinance.in believes in personalized touch so that our clients are completely satisfied with our service. In this regard, we offer One on One Customized Training to our clients.

These Trainings are provided on request by our clients & are customized according to their individual needs.

The course structure & timings for these training are highly flexible, classes are scheduled as per the convenience of our clients. 

Contact Us for More details: info@qcfinance.in.

Will be adding content and updating this post frequently.