Need maths guru

[Aussie John]

Hi this might be tricky to explain but here goes

I wish to produce a number result which is probably easiest if I use a graph as the example.

y axis (rating) steps 0 to 4

x axis (years old) 0 to infinity but realistically say 100.

I have the x axis numbers and require the y as a result. If the results were plotted the graph would look like a log curve with 0 years="4 rating" and 100 years approaching "0 rating"

25 years would represent the mid point (ie rating 2)

view example of graph

Anyone know how to code this? thanks in advance

Edited March 7, 2008 by Guest

[Søren Dyhr]

This doesn't seem too far from it...

6,5-Ln ( entry) *4,5/ Ln ( 25 )

--sd

[comment]

I think it would be helpful to know what is being computed here, i.e. the meaning of y. Because there can be several functions, of different types and therefore different curvatures, passing through the three given points.

[Aussie John]

I'm trying to condition rate a series of buildings. One of the criteria is age - ie a new building has a high rate (4) and an old building would be low (near 0) something about 25-30 years old might attract and average rating.(2)

The scale would need to be logarithmic since as a building holds its condition for a few years and and a 50 year old building may not be much more worse off that a 100 year old one (providing it is maintained).

Asides from that its more an intuitive curve rather than a pure scientific one

[Raybaudi]

My consideration:

Older building may have an highter rate than a new one, expecially if someone thinks that an old building is placed more likely in the center of the city.

The curve you are describing is more realistic a rapresentation of a car's value.

[comment]

Well, that's a pretty loose specification. I wonder if something like this would fit your purpose:

[Søren Dyhr]

expecially if someone thinks that an old building is placed more likely in the center of the city

Rome of all cities, have a bit of both - if we look beyond the 100 years, have the romans never ever suffered from nostalgia being the omnipresent decease, when it came to real estate - if something was in the way, knock it down!

This isn't math at at all, but instead a calc giving a rough or a ballpark, and as Michael says - is there an endless herd of curvatures going thru the 3 points ... but lets take it another way.

One curve might fit Farraries, and another might fit Fiats ... calculating a scrap value can't ignore the snob value to play some part in the equation. Have you heard about newer crappy buildings never made space for something even newer and in a better quality etc.

These attempts to eliminate guts feelings might be tempting, but even the largest sophisticated economic models makes from time to time wrong fortunes tellings.

--sd

[Aussie John]

This age rating is one of 10 different criteria. Also my assessment is related to the cost of refurbishing a building and not necessarily anything to do with heritage value. Regardless even worthy old buildings have a cost associated with fixing them up due to their age.

Yes it is tricky to establish the type of curve.

Edited March 6, 2008 by Guest

[Aussie John]

Comment that looks about right - And yes it is a loose specification :

What formula did you use?

Edited March 6, 2008 by Guest

[comment]

The formula is:

y = 4 * 0.0625^( x / 100 )

The 4 and the 100 are your range sizes. 0.0625 is the "magic number" that brings the first half of x's range within the first quarter of y's range, since:

0.0625^0.5 = 0.25

[Søren Dyhr]

Is it realistic to pay people to occupy a property, just because it's beyond 100 years?? Well you calc doesn't relate to your graph, the calc neatly approaches zero asymptoticly!

--sd

Edited March 6, 2008 by Guest

[comment]

you calc doesn't relate to your graph

Huh? And how do you think my graph was produced?

the calc neatly approaches zero asymptoticly!

And so does the graph.

Is it realistic to pay people to occupy a property, just because it's beyond 100 years??

What's that got to to with anything?

[Søren Dyhr]

Huh? And how do you think my graph was produced?

The tool seems to display it wrongly then, take a look!

What's that got to to with anything?

I can't get your reasoning as to why your graph leaves first quadrant at all, how would negative values play any role here?? I'm ready to admit that I reversed the issues here ... sorry!

--sd

[comment]

The tool seems to display it wrongly then, take a look!

I am looking, and I don't see anything wrong. Can you be more specific? Like, "your calc returns when x = , while your graph shows for the same value of x."

I can't get your reasoning as to why your graph leaves first quadrant at all

Because the request was for y to equal 4, when x = 0. Without the graph crossing the y axis, you will get the value of infinity for a brand new building.

how would negative values play any role here??

I don't think buildings can have negative age, but if they did, they would double in value for every 25 years back into the future.

[Søren Dyhr]

I'm mistaking the dotted horizontal line for the x-axis, what is it doing there?

--sd

[comment]

It's a cursor - drawing attention to the value of y = 0.25 returned when x = 100.

[Raybaudi]

From: http://www.walterzorn.com/grapher/grapher_e.htm

[kraai]

I'm trying to condition rate a series of buildings. One of the criteria is age - ie a new building has a high rate (4) and an old building would be low (near 0) something about 25-30 years old might attract and average rating.(2)

The scale would need to be logarithmic since as a building holds its condition for a few years and and a 50 year old building may not be much more worse off that a 100 year old one (providing it is maintained).

this is not really filemaker but...

I don't think a building's age has anything to do with its condition except as an approximation for very large sets of very average buildings. But you seem to assume the average technical life span of a building is 100 years - but as far as I know the recycle rate for office buildings is shorter, into the decades range rather than century range, because of their economic value. Even housing may be vulnerable to this. Also I can point at 40 year old buildings which are ready for demolition and 100 year old ones that are good as new. So your condition curve might actually slope upwards for older buildings, this having to do with the survival chances of buildings. The older a building is, the more likely it is to be still there in ten year's time (my hypothesis). And that would be because of its condition, which follows from its maintenance, which does have to do with heritage and market status.

So if you're working with a large dataset your curve might have meaning, but if it is smaller, take a closer look. I think you need to approach this from the maintenance and refurbishment side of things. An average building needs a new coat of paint every two to three years, new roofing tarmac every 15 years, and so on. For apartment buildings, the owner or owner collective would save up for this, and prudent homeowners might do same. In individual cases you'd make an assessment of the building's maintenance condition and look at the reservations.

Assuming you don't have specific maintenance status data for the buildings, you might try to infer a condition rating from other sources, such as building period (pre-war, post war etc, this may tell you much about construction quality and habitation quality); neighborhood, construction type (brownstone, concrete, wood), etc. and make a simple points and demerits system. Calibrate it with some experts who know the market your researching.

[Aussie John]

Thanks the basic formula is perfect for my needs. Now that I have yours and others attention I'd like to ask for another formula.

This time the curve is double curve (like a part y=cos(x) but ends continuing) but capable of being "flattened" by a variable and of course the x and Y constraints.

See Graph2 in the Graph 2 link

Thanks again

Edited March 7, 2008 by Guest

[comment]

Well it looks like arctan, but...

I am afraid you are taking an entirely wrong approach to this. As I said earlier, what really matters here is the meaning of the calculated result in real life. Once you know that, you try to build a formula that imparts said meaning to the input. The graph is merely a result of this process. If the meaning is correct, the graph too will be correct - whether it resembles your picture or not.

[Aussie John]

yes arctan is the right type of curve. thanks

In an ideal world I would take empirical numbers and get the desired curve. My result dont need to be 100% accurate - it is an estimate and as such it is easier for the curve to give the results. Adjusting the curve then gives me the results I need.

Thanks again

[comment]

Adjusting the curve then gives me the results I need.

But how can you tell that the curve needs adjusting? How do you know that the results you are getting are not the results you need? For this, you need to know something more - something that is not contained within the graph itself. Some logic that enables you to say "this isn't right".

[Aussie John]

But how can you tell that the curve needs adjusting? How do you know that the results you are getting are not the results you need? For this, you need to know something more - something that is not contained within the graph itself. Some logic that enables you to say "this isn't right".

Experience in my profession which allows me to adjust the curve.

Edited March 7, 2008 by Guest

[comment]

[Aussie John]

I entirely agree with you Comment and appreciate your advice but for me to come up with the values I would have to make an assessment (based on my experience) anyway. So yes i do have figures in my head which lead me think which curve could be appropriate. I can then test that against some examples to see how well the curve fits.

[comment]

That's OK. It's just without that information, I don't have anything valuable to add. Which is fine by me, too.

[Matthew F]

This thread is hilarious. It takes me back to algebra II.

This time the curve is double curve (like a part y=cos(x) but ends continuing) but capable of being "flattened" by a variable and of course the x and Y constraints.

How about taking a look at the 'Sigmoid Function' entry in wikipedia?

:)