Econometrics: one step ahead forecast - page 3

 
granit77:
I don't believe in tampering. Why would he do that? It's a study, not a measurement of predictions.
But you have to put a number on it to make it right.
Somehow it didn't pick up the table. Fixed it.
 
faa1947:

And there's a fudge!

Careful what you say.

We are talking about the current day and night since 00:00



1044
faa1947 09.11.2011 18:18
So the first prediction.

Using KotirOut indicator attached to the article #2 I make a sampling on D1

So: Forecast for tomorrow from 00:00 o'clock 1.3798


1044
faa1947 10.11.2011 09:33

Summary of the previous forecast.

The forecast was for a short - we have a short - the forecast is successful!

I wrote about yesterday's one for today, even though it is not yet evening (who is right), but it has already come true (in the other direction).

I will not interrupt any more.

 

How do you calculate error - is it some sort of variance? I'm sorry I didn't get that from the article, there are too many clever words.

Check your personal message.

 
faa1947:
Don't read Soviet newspapers!


https://www.youtube.com/watch?v=0mpoJh7eWjk
 
faa1947:

And there's a fudge!

Careful what you say.

We are talking about the current 24 hours from 00:00




today (where the numbers are right) is coming true.
 
ivandurak:

How do you count error - is it some sort of variance? I'm sorry I didn't get that from the article, there are too many clever words.

Look in your inbox.

I don't see it in the private line.

Considers EViews as I recall: s.c.o. + systematic error.

I do not understand the humour.

In all seriousness.

RBC websites have a consensus forecast. Five years ago I read. One, well respected Russian forecaster (I remember Maxwell Capital) predicts that Gazprom stock will fall from 320 to 250, while Merrill Lynch thinks it will go up to 400. The result: sideways for the entire forecast horizon around 320. I haven't read anything since.

 
Tantrik:

Today (where the numbers are right liyah) comes true.

Yesterday I did until 00:0 o'clock.

Today from this very 00:00 to 00:00, i.e. current day. In my post there is a table - it shows exactly everything: both the date of the fact and the date of the forecast.

 
faa1947:

Extrapolated smoothed + noise



Do you think it is sufficient that the noise is normally distributed for the adequacy of the prediction model?
 
Avals:

Do you think it is sufficient that the noise is normally distributed for the predictive model to be adequate?

The whole problem is noise, which is not normal at all - I haven't gotten to that yet
 
faa1947:

The whole problem is in the noise, which is not normal at all - I haven't gotten to that yet


not a fact if the forecast changes. Here for example let's take the normal average. Sooner or later the price will cross it again. But it doesn't mean that the price goes back to the mean. To paraphrase a popular saying: if price doesn't go to the truck, the truck will go to the price :)

For the model to be adequate, the price must return to the predicted value, rather than the predicted value moving to the price. That is, we need a mean reversion property to this "fair")) price. Then the normality of the residuals itself will be