Quote:
Originally Posted by PocketInfinities
IMO it's helpful to think of a hierarchy of tools:
Excel -> MATLAB -> Python
Use Excel for your simpler models and toying/experimenting. Use MATLAB when Excel macros don't give you enough computing power, and use Python when MATLAB won't cut it.
It's understandable that you might just want to do everything in a single language that can cover all cases (Python), but the value in MATLAB is that it's easier (imo) to code and visualize stuff than with a full-on third-generation language like Python (but of course not as easy to code and visualize as Excel). If you're new to finance you probably won't NEED to touch Python for a while.
In that hierarchy you can substitute Mathematica for MATLAB and/or C++ for Python, but I chose MATLAB+Python as I think they're more used in the industry for research purposes (compared to say, development/application purposes), which seems to be your interest per your want to self-study.
Any interest or focus within econometrics? Bayesian stuff? Micro? Macro? Volatility? The future path of interest rates?
pocket, thx for the very excellent response...
my interest in econometrics is two-fold:
1) stuff i do, time series analysis mostly and to a large degree excel is fine for coming up with the type of models i use right now - 10 data series say. maybe monthly returns for the 10 major USA industry sectors......and to some degree i don't have learn that much more for this. and i appreciate that excel is very very flexible and good for tweaking things. that was my thought before asking any of this........... i do alot of interesting work in excel myself. and i've come up with decent semi-automated analytics template with autocorrelation
2) reading other people's work........ here's where it's alot more complex and confusing given that i don't control the work/narrative. for what i've seen things like ARCH, GARCH, regime shifts etc.......... authors i've wanted to understand better are john cochrane and his collaborator monica piazzezi and others like stock/watson and diebold.................. so predicting returns and lead/lag relationships in all kinds of things....
as per #2........ what of course happens is people do all kinds of complex non-econometric math on their models i.e. setting them up, and then test the models with econometrics so that really opens the math up completely to just about anything.
one big battle is just learning the notation............ many things are understandable with the notation and things like interation functions and other ways of estimating things..... i apologize if that doesn't make complete sense.
two more things:
alot of it is just managing a large amount of data...... a large canadian money manager i'm aware of probably formally tracks thousands of indicators and has done very small studies on most of them i.e. you read about comstock indicator so you program it, doing a cursory back-test and make a few comments and then save it.. then anyone can access it, they can call it up at any time. they might even do PCA on it with other similar indicators to try a come up a with a purified indicator for it. ........ so some of my interest is just learning computer skills to set something up like that.