SERVING THE QUANTITATIVE FINANCE COMMUNITY

alinastef
Topic Author
Posts: 12
Joined: July 18th, 2008, 6:39 pm

### Fitting a Student's t distribution

Hello,I'm trying to fit a Student's t distribution to the data (because I have fatter tails), how can I do it and obtain the number of degrees of freedom? Matlab offers normfit for fitting a normal distribution, and it also offers fits for many other distributions but not for Student's t. Thanks.

Alan
Posts: 10318
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

### Fitting a Student's t distribution

alinastef
Topic Author
Posts: 12
Joined: July 18th, 2008, 6:39 pm

### Fitting a Student's t distribution

Thanks.

ronm
Posts: 163
Joined: June 8th, 2007, 9:00 am

### Fitting a Student's t distribution

for estimating d.f., MLE may not be good approach. If you are really sure the DGP is some t distribution then you may try with standard statistical tests applied to infer the population distribution like KS test etc, with different d.f. values. Otherwise you can check graphically with Q-Q plot etc.However please be warned that those tests are just for rejection, not really for acceptance.Thanks,

Alan
Posts: 10318
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

### Fitting a Student's t distribution

QuoteOriginally posted by: ronmfor estimating d.f., MLE may not be good approachwhy not?

yuryr
Posts: 402
Joined: November 5th, 2007, 12:47 pm

### Fitting a Student's t distribution

Student-t can be fitted using EM iterative algorithm (which is maximum likelihood) Look at this paper for example.This distribution is more stable with regards to outliers (unlike Normal...).

list
Posts: 2041
Joined: October 26th, 2005, 2:08 pm

### Fitting a Student's t distribution

Student-t compare Gaussian model distribution with a histogram of a given sample observation. From the common sense more degrees of freedom is better than less. From my memory in old statistics books that studied before software replaced hand calculations it was recommended that in each cell should be more than 30 observations. This somewhat guarantees reliability of the test.

ACD
Posts: 107
Joined: April 19th, 2004, 8:09 am

### Fitting a Student's t distribution

QuoteOriginally posted by: listStudent-t compare Gaussian model distribution with a histogram of a given sample observation. From the common sense more degrees of freedom is better than less. From my memory in old statistics books that studied before software replaced hand calculations it was recommended that in each cell should be more than 30 observations. This somewhat guarantees reliability of the test.I think you're referring to the t-test rather fitting a student-t distribution.

list
Posts: 2041
Joined: October 26th, 2005, 2:08 pm

### Fitting a Student's t distribution

QuoteI think you're referring to the t-test rather fitting a student-t distribution.Sorry, you are right

Wilmott.com has been "Serving the Quantitative Finance Community" since 2001. Continued...

 JOBS BOARD

Looking for a quant job, risk, algo trading,...? Browse jobs here...

GZIP: On