• C++ Programming for Financial Engineering
    Highly recommended by thousands of MFE students. Covers essential C++ topics with applications to financial engineering. Learn more Join!
    Python for Finance with Intro to Data Science
    Gain practical understanding of Python to read, understand, and write professional Python code for your first day on the job. Learn more Join!
    An Intuition-Based Options Primer for FE
    Ideal for entry level positions interviews and graduate studies, specializing in options trading arbitrage and options valuation models. Learn more Join!

Prob. Distribution

mercury

Joined
7/27/10
Messages
13
Points
11
Guys I need a little help. I'm sure someone else has already seen this problem.

The normal distribution does not fit the financial mrkts well: tails are too thin (it's not uncommon to see moves beyond 3, 4 and more ST. Deviations), if the probab. distrib. has skewness would be great!

Thanks in advance to everybody :)
 
What exactly are you looking for? You can skew the normal distribution, and even add kurtosis if needed. You can pick a different distribution to work with as well, Student's t and any other power law distributions all have fatter tails which may work depending on your data.
 
i could be completely wrong here so just take this as an idea. you are stating that the normal distribution's tails are too thin. could you use the t-distribution? it is very similar to normal, however, its tails are thicker. they say to use t-distribution when CLT theorem doesn't apply (when you have <40 samples), however, i don't think that rule is relevant in your case. the tails are thicker to adjust for variations of accuracy of data due to small sample sizes.

... something you could look into. i'm not an expert - i'm just taking prob & stats right now so it's semi-fresh in my brain.

Tylor
 
Guys I want to thank all of you for your inputs. They are all great. My problem (or at least 1 of them) is that moves with 3, 4 and at times more STD, are recorded so I'm I'm trying to leave the normal (and lognorm) distr. as last option. If I can find another f() that would be great.

Does any of you has experience on how to skew a t-student.

and for the Levy Distr:
- is it based on brownian motions?
-is there a paper out there that shows the chart and the various shapes for different parameters?

Tnx again guys - much appreciated!
 
U can make the normal distribution tails thinner if you want...and add skewness and kurtosis as desired(of course by taking certain restrictions into consideration). You make median of the data more or less in value (by adding it) than mean creating a positive or negative skewness. If you want to leave the symmetry of the data as it is (if perfectly or nearly perfectly symmetric) than you can make its tails thinner to fit the financial reasoning. and lastly why dont try to use a t-dist or modified logistic distribution?!!
 
Thanks a lot David!
t-dist or modified logistic distribution I believe are better (normal is too thin in terms of tails and too symmetric).
Again: thanks a lot Man!
 
I would send you a family of distributions u might find useful if u have skype or yahoo or mail where i can send. By the way, when evaluating financial data that can logically be thought of as exposed to large tails be careful in using modified logistic distribution for continuous X-s. ;) Tails can only be expanded when making sure u have enough n observations
 
modified logistic distribution
Click to expand...

Which is the modified logistic distribution??? As I know there is only the logistic distribution. What is modified???? Interesting ... :)
 
Back
Top