Econometrics: one step ahead forecast - page 76
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
You can't remember them, because they don't exist. It would be too easy to make money in the market then...
If we are talking about mathematical proof, my lack of knowledge is not proof and I would not generalise.
What do you have a problem with my (worldwide) empirical technique?
It's the lack of formal evidence that doesn't satisfy you.
Yes, you realise this yourself, as you strongly doubt your model.
P.S. No finite set of tests on a finite set of data can provide sufficient conditions for prediction.
"The finite data set" is also in some way a statistic, defined by the proximity of the test statistic under investigation to the marginal statistic. In other words, depending on the level of significance, you can sometimes consider the amount of data to be finite or infinite. This is a load of bollocks, of course, but it seems so to me.
faa1947:
That fitting is a bad thing is the opinion of this forum. Any student around the world who has taken a course in econometrics or statistics does not think so.
We don't care what the adherents of the econometric sect, who have been brainwashed with mathematical tricks, think. Because we are only interested in how the broker counts our money. And brokers don't pay money for tweaks.
You have the wrong forum and if you want sympathy and understanding, look for it amongst followers of the econometric religion just like you.
Just don't bring statistics into this. The concepts of stationarity in statistics and econometrics do not correspond to each other. In statistics, stationarity is independent of the sample; in econometrics, "stationarity" is the result fitted to a particular sample.
In terms of mathematical proof, my not knowing is not proof and I wouldn't generalise.
What do you have a problem with my (worldwide) empirical reception?
and are you familiar with cointegration? This concept and the tests are somewhat broader than just the stationarity of the residuals. I.e. the price must be cointegrated with your regression. For that, their linear combination must be stationary, which is what you test (residuals). But apart from that, the "error correction model" is also evaluated.
Prior to the Engle-Granger method, researchers often unknowingly obtained false regressions, or estimated regressions in new differences, which, although leading to stationarity of the variables, did not allow for the stationary correction term, i.e. the regression model was incorrectly specified (omitted variable problem). This underlines the role of the correction term (it is assumed that if in the previous period variable Y deviated from its long-run value, the term
corrects the dynamics in the right direction). http://ecnmx.ru/article/a-97.html
The way I see it is a return to a predicted value
more link))) h ttp://www.nsu.ru/ef/tsy/ecmr/coint/green/green184.htm
We don't care what the adherents of the econometric sect, who have been brainwashed with mathematical tricks, think. Because we are only interested in how the broker counts our money. And brokers don't pay money for tweaks.
You have the wrong forum and if you want sympathy and understanding, look for it amongst followers of the econometric religion like you.
Just don't bring statistics into this. The concepts of stationarity in statistics and econometrics do not correspond to each other. In statistics, stationarity is independent of the sample; in econometrics, "stationarity" is the result fitted to a particular sample.
I haven't made a mistake with the forum. I am well aware that fooling the heads of people who come to the field of miracles with TA and NS will be more difficult, but I will participate and wait for the ban.
And what miracles do you see in TA?
TA merely implies the dependence of subsequent prices on previous prices. Is that a miracle?
I did not make a mistake with the forum. I am well aware that fooling the heads of people who have come to the field of miracles with TA and NS will be more difficult, but I will participate and wait for the ban.
You won't get a ban, we won't make a hero out of you.
Speaking of NS: there is a quite reasonable empirical technique in decent nerve packs to test NS for generalisation ability, i.e. prediction ability (stopping learning when the minimum error on the verifying dataset is reached). And for some reason I think it is more scientific than your arbitrary set of tests.
Yes, you know that yourself, because you have great doubts about your model.
I don't doubt it. It seems to me that it is possible to find a solution to an extremely limited problem: build a robust model for a sample of 1 more. Not in general past and future, but only +1. My model just shines inside the sample, what's wrong with it that outside the sample it leaks at all (you can get a profit factor of up to 2)?
"The final data set" is also in some way a statistic, defined by the proximity of the test statistic being examined to the marginal statistic.
Should we understand your answer to mean that by not solving for n out-of-sample, we won't solve for +1 out-of-sample?
Why are we arguing about marginal sampling at all? Let us solve this question restricted to +1
You're talking about the wrong sample. I'm talking about the raw data.
P.S. In fact, I still don't understand why the null test you insist on for regression coefficients is still not applied to the forecast? You have already had the magnitude of the predicted price change several times much smaller than the prediction error. Yet you persist in posting such a forecast. Is that what you call a scientific approach?
And what wonders do you see in TA?
TA merely implies that subsequent prices depend on previous prices. Is it a miracle?