Test significance parameters mvnrmle regression

Hi to all,

I'm trying to figure out how to estimate the t-score for the parameters fitted by the function mvnrmle. I have to perform a factor model analysis, the Fama-French three factor approach in particular. The only script I found on Internet was this one that applies the mvnrmle function and not the Fama-Mac Beth two stages function. Anyway, could someone tell me how to test the parameters estimated? I don't know how to get the betas' standard errors.

Thanks to all

Answer 1

[Brendan Hamm]

There is really no reason to be using mvnrmle to fit this model. If you are using 2013b of later I would suggest the use of fitlm since really the French-Fama model is just a linear regression model. This then has methods such as coefTest which will allow you to perform the appropriate F-test on the coefficients.

Answer 2

[Antonio Amoretti]

Thank you Brendan! Actually I used that script because I'm quite new to Matlab, so I follow the author. Once I run the fitlm, is there a second step as done in Fama-Mac Beth to do to complete the regression? I have 200 monthly observations for 43 stocks, and I've to apply the fama French model with regressors the data from their website. Thank you a lot

[Brendan Hamm]

What is the source for the code you are using and the data you have? Were you looking to perform the Fama-MacBeth 2-stage regression or just simply to apply the French-Fama 3-Factor model?

Answer 3

[Antonio Amoretti]

I downloaded STOXX50 costintuents daily and monthly returns from Datastream. I initally used daily returns running that mvnrmle regression on generic three factors daily data (FF website). The results were not satysfying when adding the SMB and HML factors, so I used the regional european factors ( available only monthly, so I used monthly returns for assets). My aim is to implement a factor investing strategy and allocate the same risk budget to each factor included in the factor model analysis: for each month, allocate the same risk budget to the factors considered such that each of them ( Market-rf, SMB, HML and eventually Momentum factor) contributes in the same way to the total variance in the portfolio. So my first step was to compute the factor loadings and then to decompose portfolio risk respect to these factors and successively implement the risk factor strategy. So I actually don't think I need to run the two stage Fama-MacBeth.

What could you advice to me?

Thank you

[Brendan Hamm]

This forum is meant for help with MATLAB questions and it seems that this question is now venturing into investment strategy help. This is not the appropriate place for such questions. If you require assistance with actual MATLAB code as it relates to this problem, people will assist. Without any posted code to work with it is not possible to provide further assistance.

Also, please Comment using the "comment on this Answer button" and not in the "Answer this question" section.