Econometrics: one step ahead forecast - page 92
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
I have to make sure that the "flow of losing trades" is stationary.
Personally, it is more important to me that the "flow" of losing trades is stationary.
If the TS has made $10 in the first month, $20 in the second, $40 in the third, $80 in the fourth, etc., then this series is non-stationary and therefore the profit is random?
If the TC brought in $10 in the first month, $20 in the second, $40 in the third, $80 in the fourth, etc., then this series is non-stationary and hence this profit is random?
You can only predict the trend, you cannot predict the noise. I believe that cotier = trend + noise. I model the trend in analytical form. There is no non-stationarity at all here - it is a deterministic thing. All non-stationarity is left in the noise, the residual. There is no non-stationarity anywhere else, to the right.
Econometrics is not a science. It is an applied discipline which studies the application of matstatistical methods and theorists to economics. There are no special methods, models and apparatus in econometrics.
All methods of matstatistics without exception are based on the hypothesis "deterministic residual + stochastic residual". "Trend" is the deterministic residual of a time series. Almost all matstatistics methods are created only and exclusively for stationary and ergodic series.
If a numerical series is non-stationary, then this non-stationarity cannot be "left in the noise".
In order to apply most of matstatistics methods, the series should be stationary, ergodic and its residual in a model should have normal distribution.
Farnsworth has suspiciously straight function lines that do not reflect changes in the trend.
Fansworth's lines are fine.
There are 270 observations here. If we draw the ACF, we can see a wave that corresponds to the change in trend. The 270 bar does not fit in the screen, so I will give the part where the trend changes from 40 to 90. This is what it looks like:
Once again, ACF quotes are meaningless to watch, do you understand that? It is not a stationary process. If you want to repeat the experiment, why 270 bars? Take a sample of 5000 (can you feel the difference between 5000 and 270?).
We can see the ACF wave matching the visual trend. The instructions in EViews warn us to be careful when specifying the number of lags on which we count the ACF. For anyone familiar with TA, this is a known problem, as it is possible to draw many trends on one plot.
Wow, what do you see there? Tell me, what do you use to expand your consciousness? Is it some kind of plant extract? What do you use, cacti?
About 40 years ago Box and Jenkins wrote a book about ARMA, 300 pages long - you think too much of me that I can do it on this forum in a nutshell.
it's a very good book and please don't "explain" it, you're really bad at it.
to Trolls
Take 500 counts and plot the ACF. ("... The autocorrelation function plot can be obtained by plotting along the ordinate axis the correlation coefficient of two functions (base and function shifted by the value τ), and along the abscissa axis the value of τ ...")
The entire sample is 5000 samples (read carefully), and I looked at the correlation for the first 500 samples. Trolls, - the calculation is correct. For other variations of length and time interval the ACF will be different, as you and our econometrician have shown. don't worry, keep yourself busy with something useful.
If the TC brought in $10 in the first month, $20 in the second, $40 in the third, $80 in the fourth, etc., then this series is non-stationary and hence this profit is random?
do i look like a psychic? how do i know if the returns of your system are stationary or non-stationary using this data? What do months and dollar returns have to do with it at all?
in any case, it is desirable that system returns are close to stationary and match test returns. Especially in a series of trades. If you initially believe that this is not the case, then you cannot trust the results of historical tests - you can just disregard them. I.e. quasi-stationarity of results of a series of trades is still needed.
I disagree here. If the results are non-stationary, it means that the used measuring tools like s.c.o. only approximate the process and are not accurate. This does not mean that the process cannot be profitable. The problem is only with the tools we use to measure the process itself, but not with the process.
Econometrics is not a science. It is an applied discipline which studies the application of matstatistical methods and theorists to economics. There are no special methods, models and apparatus in econometrics.
All methods of matstatistics without exception are based on the hypothesis "deterministic residual + stochastic residual". "Trend" is the deterministic residual of a time series. Almost all methods of matstatistics are created only and exclusively for stationary and ergodic series.
To apply most matstatistics methods, a series must be stationary, ergodic and its residual in the model must have normal distribution.
Practically all methods of matstatistics are created only and exclusively for stationary and ergodic series.
To apply most of matstatistics methods the series must be stationary, ergodic and its residual in a model must have normal distribution.
Big news. WHAT ABOUT ARIMA? WHAT ABOUT ARCH? What kind of series are these for?
If the numerical series is non-stationary, then this non-stationarity cannot be "left in the noise".
cotier = trend + noise
On the left is non-stationarity, but where on the right?
Wow, what-what-what do you see there? Tell me, what are you expanding your consciousness with? Is it some kind of plant extract? What do you use - cacti?
I disagree here. If the results are not stationary, it simply suggests that the measurement tools used, such as the same s.c.o., are only roughly describing the process and are not accurate. This does not mean that the process cannot be profitable. The problem is only with the tools we use to measure the process itself, but not with the process.
It's about the results of the system. If they are non-stationary, then the calculated mo, sko, etc. may be quite different in the future. You get a Mo of +10 points on tests, and in the future it's not the same at all - the range is huge and +10 is not its average value. It is clear that stationarity is too rigid and abstract a condition - we are talking about quasi-stationarity - statistical parameters change quite slowly. I.e. dividing everything into stationary and non-stationary is like black and white. There is a compromise and many shades in between.)
You're a boor.
No, not a boor, but a very cultured and tactful person. This pathological bullshit of yours, when you visually try on the ACF chart to the price chart and so on ... is just a bit annoying. And like a true econometrician (for what I "adore" them) instead of M15 you take H1, instead of 5000 points you take 270 and cheekily puffing out your words like "the lines are different... it's suspicious, like...".
All right, you literate man, if you have offended me, I apologize, I was hoping that you would not reply to my posts. You do not answer them, let's part as econometricians - without saying goodbye.
Good luck and good trends.