Practical advice, please. - page 6

 
Alexey Viktorov:
And what do ArrayMaximum and ArrayMinimum not like? Why did you need to write it through a loop?

I was expecting this question ))
I decided to write my own functions to understand how they work.
And at any moment I can adjust them to the task.
Also these Mql functions return index of found element, not value.
I consider extra dancing with value definition as unnecessary.

 
Roman:

I was expecting this question ))
I decided to write my own functions to understand how they work.
And at any moment I can adjust them to the task.
Also these Mql functions return index of found element, not value.
I think that additional dancing with value definition is unnecessary.

This is not far from writing my own MT.

ps; And you are wrong in thinking that finding a value from an index is a "dance". Having an index of found element of array we can easily pass to index of bar if necessary, but finding index by value is really a tambourine dance. But you know best. Only in vain you are trying to advertise yourself and your handicrafts.
 
Alexey Viktorov:

It's not so far-fetched as to write your own MT.

ps; And you are mistaken that finding a value by index is "dancing". Having an index of found element of array we easily pass to an index of bar at necessity, but search of an index by value is really a tambourine dance. But you know best. Only in vain you are trying to advertise yourself and your handicrafts.

What's the point of advertising? If only there was something...
I needed values, so I took them and made them for the task.
What is being done for nothing or not is irrelevant.

 

Good evening to all. )))

Dimitri, your rating system is understandable, but it doesn't take into account the spread that you yourself realised and wrote about -"Just by the average isn't enough, you also need to have no large outliers. "

Raman, your approach to the problem is not obvious to me personally... And if I don't understand it, my attitude is quite sceptical ))))


Now for what I've dug up.

Nah, tomorrow. Tired .... )))

 
Сергей Таболин:

Good evening to all. )))

Dimitri, your rating system is understandable, but it doesn't take into account the spread that you yourself realised and wrote about -"Just by the average isn't enough, you also need to have no large outliers. "

Raman, your approach to the problem is not obvious to me personally... And if I don't understand it, my attitude is quite sceptical ))))


Now for what I've dug up.

Nah, tomorrow. Tired .... )))

Well if you don't understand, then read the literature
Since your errors are expressed in percentages, MAPE will probably do.
Tweak the function on the first page and you have MAPE.

Основные оценки точности прогнозирования временных рядов
Основные оценки точности прогнозирования временных рядов
  • www.mbureau.ru
Работая с научными публикациями, сталкиваюсь с различными показателями ошибок прогнозирования временных рядов . Среди всех встречающихся оценок ошибки прогнозирования стоит отметить две, которые в настоящее время, являются самыми популярными: MAE и MAPE . Пусть ошибка есть разность: , где Z(t) – фактическое значение временного ряда, а –...
 
Maybe this will also help.
Прогнозирование временных рядов при помощи экспоненциального сглаживания (окончание)
Прогнозирование временных рядов при помощи экспоненциального сглаживания (окончание)
  • www.mql5.com
В статье "Прогнозирование временных рядов при помощи экспоненциального сглаживания" [1] были кратко представлены модели экспоненциального сглаживания, продемонстрирован один из возможных подходов к оптимизации параметров моделей и в конечном итоге создан индикатор, производящий прогнозирование на основе модели линейного тренда с демпфированием...
 

So...

I've been reading and thinking, and I've decided that there are four things I should be interested in:

According to the data given, these values look like this:

STANDOTCLONE 7,8208
7,9133
8,4150
MEDIANA 6,3300
6,3300
5,0600
EXCESS 1,1322
1,9702
1,1832
STANDARD ERROR 2,02
2,04
2,17


Question: how do I put this data together already? I have thought of it this way:

First option.

STANDARD SLOPE - MEDIAN + KURTOSIS

2,62
3,55
4,54

divided by STANDARD ERROR

1,30
1,74
2,09


Second option.

Average error/STANDARD ERROR - Average error/MEDIAN + Average error/EXCESS

7,77
4,14
6,89

divided by STANDARD ERROR

3,85
2,02
3,17


What do you think of this?

 

Sergey, I apologise if I'm off-topic, I haven't quite figured out what you're doing. But I may have something useful to say about your last post.

When I tried to replace two numbers with, say, the difference or quotient of these two numbers, some information is lost. So I'm not sure about the need to do mathematical operations on the indicators "all grit". It is better to compare them to each other in pairs.

There are some connections between the different measures of descriptive statistics. If the mean is almost equal to the median, it means that there are no outliers in the sample, because the median is resistant to outliers, while the mean is not. If the mode is shifted away from the mean, it means that the distribution density is asymmetric. Then there are some more.

 
Aleksei Stepanenko:

Sergey, I apologise if I'm off-topic, I haven't quite figured out what you're doing. But I may have something useful to say about your last post.

When I tried to replace two numbers by, say, the difference or quotient of these two numbers, some information is lost. So I'm not sure about the need to do mathematical operations on the "whole gurt" numbers. It is better to compare them to each other in pairs.

There are some connections between the different measures of descriptive statistics. If the mean is almost equal to the median, it means that there are no outliers in the sample, because the median is resistant to outliers and the mean is not. If the mode is shifted away from the mean, it means that the distribution density is asymmetric. There are also some.

The opposite only.

 
Aleksey Mavrin:

Just the other way round.

That seems about right.