# Tag Info

4

Keep in mind that Benford's law is not a universal or natural law. A violation of Benford's law is neither a necessary nor a sufficient condition to prove a flaw or a quality issue in the data. At the best, it can give you a hint, but it should not be trusted blindly. Moreover, note that for some types of data the law will not work at all, such as e.g Likert ...

3

I would argue that indeed none of the so-called stylized facts you mentioned can be explained by classical economic theory. That there was a gross delta between the predictions of classical economic theory and empirical data was foremost found out by Benoit Mandelbrot as far back as 1963 in his seminal paper: The Variation of Certain Speculative Prices In ...

3

2) Alternative to Fama-MacBeth is Fama-French approach. Explanation of difference see, for example, here: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1271935 Fama-French approach was used by Carhart (introduced momentum), Pastor-Stambaugh (introduced liquidity), Fama-French themselves (used it to build 5-factor model), and many other (elsevier or ...

3

PX_BID and PX_ASK are the static equivalents of BID and ASK, the latter two of which populate in "real time" (i.e. as they are dynamically updated). So the PX_BID and PX_ASK values are dependent upon when you pulled the data. Bloomberg's source depends on the asset in question and the exchange on which they are listed, but the data does come from the ...

3

As with many machine learning technologies, you can run a separate training and testing phase before deploying it live for prediction. All it does is build a collection of decision trees based on the parameters you give it - if the output field is a factor, you get classification (a finite enumerated set of values); if it's numeric, you get prediction. One ...

2

You should de-trend to whatever frequency scale you are testing. I.e. 1 min means de-trend 1 min data. Merely by moving to higher frequency data, you are eliminating much of the systematic bias present at higher scales -- as 1) you have many more samples to compare (minimizing standard error) 2) At smaller intervals, the drift component also shrinks ...

2

I have not used random forests myself but I know of a guy who applied this classification technique to machine learning algorithms applied to pattern recognition. Thus I think its advantages over classic regression approaches can be applied to discern patterns in financial data, though I get the impression that it vastly overfits the data and thus you end ...

2

The key assumption is that there is no time-series correlation between the error terms. Fama-MacBeth can deal with cross-sectional correlations. See Samuel Thompson's "Simple formulas for standard errors that cluster by both firm and time" in the Journal of Financial Economics (2011) for a treatment of different regression methods for testing equity ...

2

I think there is a slight misconception into the purpose of an economic theory. The market is a complex entity to be modeled and yes, it is neither efficient nor arbitrage free but it is trading and there is a price process that corresponds to the market one. You could say that classical economic theory has failed, but I would argue the idea of a theory is ...

2

They represent the current BID and ASK at the time you query them. If you look up those fields in the terminal FLDS<GO> you will see they are marked as reference data, that means they are not continually updated. They are refreshed each time you query them. They come from the NBBO quote at the time you query them.

2

For the general solution in the case where $f$ is not a constant, note that, from the SDE \begin{align*} dx_t = \theta(f(t)-x_t)dt + \sigma dW_t, \end{align*} we obtain that \begin{align*} d\big(e^{\theta t} x_t \big) = \theta e^{\theta t} f(t)dt + \sigma e^{\theta t} dW_t. \end{align*} Then \begin{align*} e^{\theta t} x_t = x_0 + \int_0^t \theta e^{\theta ...

2

You can just take expectations on both sides of your SDE/corresponding integral equation and obtain an ODE on the expectation function $m_t = \Bbb E[x_t]$: $$\dot m = \theta(f - m)$$ which you can easily solve using ansatz $m_t = c_t \mathrm e^{-\theta t}$ which brings you to  m_t = x_0\mathrm e^{-\theta t} + \theta\cdot\int_0^tf(s)\mathrm ...

2

I think the only valid answer is you can't. The techniques you describe would work of the signal was much stronger than the noise but it seems that with your fund returns this is not the case. You could try to get more data or look at other risk measures like max drawdown to get some idea of the risks involved.

1

The rating downgrade/upgrade effect is definitely more extreme during financial crisis, because of several effects (among all, flight-to quality, flight-to-liquidity and news effects itself), as shown by: Arezki, Rabah, Bertrand Candelon, and Amadou Nicolas Racine Sy. "Sovereign rating news and financial markets spillovers: Evidence from the European ...

1

According to me, you should be consider the use of the fractal distribuion/law power distribution in risk management. Currently, those topics are up-to-date in the risk management area and, more generally, in finance since those probability distribution should predict financial risks better than actually the Normal distribution do (see, e.g., the fat-tails ...

1

50 elements input vector is actually a small one. For example, in this tutorial the size of the input vector is 784 (parameter 'nvis'). So your problem lies somewhere else. I would recommend to start from taking these two courses on Coursera: Neural Networks for Machine Learning Machine Learning They will provide you with some practical guidance ...

1

Yes and no :-) Portfolio VaR = CV1 + CV2 + CV3 + CV4 is correct. To safeguard my answer, I looked this up from thinxlabs.com The individual component VaRs from the assets in the portfolio should add up tho the total portfolio VaR. The equation is as follows. But you need to calculate another VaR for each account, if you want to use CV on those. The ...

1

Classical economics cannot "explain" volatility smiles, but neither does it preclude their existence. Economics is far more abstract than financial "quant"modeling and answers very different questions. In the more abstract framework of economics, volatility skew, mean reverting volatility, bubbles, and crashes are all conceivable scenarios. ...

1

You said that the PX_BID and PX_ASK values are dependent upon when you pulled the data. But if I pull historical data (e.g. for the last month) only the value of today would be changing but not the past ones... So there should be a point of time when the final PX_BID and PX_ASK values for a day are calculated. Or am I wrong?

1

Recently I found a book on earnings trading but did not have time to read thoroughly. Trading on Corporate Earnings News - John Shon I also had spent some time to see earnings surprise effects and it is a quite interesting but not easy to use topic. There is certainly a jump if the estimates and announced earnings have a large mismatch but the magnitude ...

1

I agree with @MattWolf The graph you show is confusing and evil, it makes me feel dumb every time I look at it. So I inverted the axis. Now we see the familiar shape of an utility curve, discussed in your previous question. It is upward sloping at a declining rate. In this case $u$ takes the place of $R_p$ and the general form of mean variance utility is ...

1

It could help with things like fraud detection, analysis of bankruptcy probability, default risk, unsupervised learning for qualitative/descriptive purposes, or for a purely backwards looking supervised analysis on returns again for descriptive/understanding purposes (variable important, etc, perhaps impulse response analysis). It may also be good at ...

1

In answer to your question 2, you should detrend over the entire range of the back test period. The purpose of the detrending is to satisfy/create the null hypothesis for the boot strap test (it's not strictly necessary for the permutation test). This hypothesis is that the return from your strategy is zero. To create this zero null hypthesis you have to 1) ...

1

I think the simplest way to achieve what you're looking for is through regression coefficient hypothesis testing. Perform linear regression on returns (y-axis) vs. dates (x-axis) over the desired time frames (do it once for 5 months, once for dataset w/15 months worth of data, and once for 60 months worth of data). As a result of regression, you will get ...

Only top voted, non community-wiki answers of a minimum length are eligible