Dear community,
For my Master Thesis, I am planning to replicate the Fama French 5 factor model from scratch. However, I encounter a practical problem in constructing the factors.
In order to construct, for example, the factor SMB, I need to divide my dataset in two parts with respect to size and in three parts with respect to value/growth stocks (as explained in your papers). My question is how do you do this? The dataset I uses consists of approximately 8000 companies with a range of approximately 50 years in panel-data format. What techniques should I use in order to create these factors? And should I do this in panel-data format or time-series format? I thought using time-series format (i.e. one column for every firm) I might be able to make the distinctions since every firm had one own specific variable. However, unfortunately, my dataset is too large in order to convert it into time-series format (even when I try to divide it in subsamples). Using panel-data however, all firms are listen underneath one variable and hence I struggle with the question on how to divide these firms into different groups in order to create the factors.
I hope you can help me with my problems!
Yours truly,
Niek van der Schaaf
You need to divide your dataset in two parts with respect to size and in three parts with respect to value/growth stocks , but EACH June. So, using a panel should not be a problem for you
Thank you for your answer Miguel Angel Martinez Sedano , but what techniques are needed in order to divide the dataset? In other words, how do you divide the dataset when it is in panel format?
Year by year, June by June
Build the six portfolios en each June and maintain it till the next June
Dear Miguel Angel Martinez Sedano thank you for your answer! However, I meant what statistical techniques are needed for this procedure? Since I work with approximately 8000 firms, doing this manually is not really an option unfortunately. I use the program Stata. Are you known with this program?
Unfortunately, I am not a user of Stata, but I think it is not and imposible task. You must program Stata carefuly; I do not know any statistical method to solve your problem
- Badges
- Science topic
- Similar topics
- Papers