Correct calculation of currency indices. - page 21

 
faa1947:

Healthy people derive their knowledge from their own observations and reasoning, not from the MICEX website.

I want to make you happy.

Most of you cyclists are like that.

The number of cyclists here is one and two.

And it seems to you that there are a lot of us, simply because you are afraid of us - for our stupidity is too obvious for us, and you slightly guess about it. :)

Hence your regular aggression towards independent thinkers - a kind of active-defensive reaction. Yeah, I probably shouldn't go on or I'll get banned.

Why don't you go and read a book, it'll calm you down.

 
MetaDriver:

Why don't you go and read a book, it will comfort you and calm you down.

Good advice, because the impression is that the comrade is not even reading books, but manuals for packages...

(

____

going back to the index - should the index be required to include as many currencies as possible?

i.e. what is the optimal number?

5-8 or covering 95% of world GDP by the issuers in the index?

;)

 
avatara:

1. going back to the index - should it be required to include as many currencies as possible?

2... i.e. what is the optimal number? 5-8 or covering the index member countries with 95% of world GDP?

;)

1. it shouldn't be required, but you can dream about it. :)

2. I purely practically choose a compromise between accuracy and calculation overheads and use just 5 to 8. In principle, the bigger the basket, the more accurate you can calculate. But the time consumption for calculations grows quadratically. Another problem is the spread. It introduces uncertainty into calculations, which linearly grows with growth of basket, though practice shows that if you take an average (Bid+Ask)/2, the errors will not be too big (arbitrage rules, it keeps things close). Evenmore uncertainty brings the way of storing quotes in MT: the minimum discreteness is minutes, and inside minutes we do not know when Low and High... :) Additional expenses are required for calculation pairs (such as XAUGBP or NZDXAG). I am not even speaking about the necessity of pre-synchronisation and closing of quotation gaps...

;)

 
MetaDriver:


Everything seems to make sense with prices, but how can the volume of the index be calculated correctly?

 
BoraBo:

Everything seems to make sense with prices, but how can the volume of the index be calculated correctly?

Honestly, I don't know. I don't even try to calculate correctly, i.e. to estimate the real volume based on the tick volume, taking into account the index value of the asset relative to the denominator of the basket. That would still be too many assumptions. I simply sum up all the tick volumes of the pairs for which the corresponding index is calculated, and for the calculated pairs I take the arithmetic mean of the initial pairs. It is obviously mathematically incorrect, but for my purposes (early diagnosis of the beginning of movement of a particular currency) is enough.
 
MetaDriver:

2. In purely practical terms, I choose a compromise between accuracy and calculation overheads and use just 5 to 8. In principle, the bigger the basket, the more accurate you can calculate. But the time consumption for calculations grows quadratically. Another problem is the spread. It introduces uncertainty into calculations, which linearly grows with growth of basket, though practice shows that if you take an average (Bid+Ask)/2, the errors will not be too big (arbitrage rules, it keeps things close). Evenmore uncertainty is brought by the way of storing quotes in MT: the minimum discreteness - minutes, and inside minutes we do not know when Low and High... :) Additional expenses are required for calculation pairs (such as XAUGBP or NZDXAG). I am not even speaking about the necessity of preliminary synchronization and closing of quotation gaps...

I see, I'm stuck with technical problems. Yes, we know, we're trying to swim.

What kind of visualization do you have? In short, see my personal message.

 
MetaDriver:
Honestly, I don't know. I don't even try to calculate correctly, i.e. estimate the real volume from the tick volume, taking into account the index value of the asset relative to the denominator of the basket. That would still be too many assumptions. I simply sum up all the tick volumes of the pairs for which the corresponding index is calculated, and for the calculated pairs I take the arithmetic mean of the initial pairs. It is obviously mathematically incorrect, but for my purposes (early diagnosis of the beginning of movement of a particular currency) is enough.

That's how I calculate volumes too. But suppose we get real volumes (e.g. from Dukas), then is it just the arithmetic average too ?

 
BoraBo:

That's how I calculate volumes too. But suppose we get real volumes (e.g. from Dukas), then only the arithmetic average too ?

No. Then I would recalculate the volumes in the currency of deposit, then sum it up, and at the end one of two things: either leave it in the currency of deposit (which is practical, because it allows to compare instantly), or recalculate it in the currency of index, through the rate of currency index to the currency of deposit, which is more academic, but, imho, impractical. For calculated rates all the same I would have to make some arithmetic average assumptions. Something like this.
 

Calculus is kind of a maths term.

Everything in calculus begins with a properly defined problem.

The problem is...

There is a closed system with 8 variables({"#EUR", "#USD", "#GBP", "#AUD", "#NZD", "#JPY", "#CAD", "#CHF" } whichever is bigger).

Given 7 inverse functions to the second variable .

Find the values of 8 variables and the error of calculation.

//--

Do I understand the problem correctly?

And why volumes, because for currencies do not play a role (except for people's activity).

 
MetaDriver:
Nah. Then I would recalculate the volumes in the currency of deposit, then sum them up, and at the end one of two things: either leave it in the currency of deposit (which is practical, because it allows to compare instantly), or recalculate it in the currency of index, through the rate of index currency to deposit currency, which is more academic, but, imho, impractical. For calculated rates I would still need to make some arithmetic average assumptions. Something like this.

Yep, that's right (I was too slow), that's what you should do, bring all volumes to a single currency, it doesn't have to be a deposit currency, just one currency and you can do whatever you want. You may do whatever you want. Thank you.

I would not go anywhere without averaging assumptions when constructing my own indices.