# Fractions as approximations

• Mar 18th 2007, 10:51 PM
slobone
Fractions as approximations
The other day I heard the business reporter on the radio say "Advancing stocks led declining stocks by 7 to 5." That is, 7/12 of all stocks went up for the day, and 5/12 went down.

Now, there are thousands of stocks in the market, so 7/12 is most likely just an approximation to the actual number. My question is, how good an approximation is it? You have enough information to answer the question if you make one reasonable assumption.
• Mar 18th 2007, 11:12 PM
topsquark
Quote:

Originally Posted by slobone
The other day I heard the business reporter on the radio say "Advancing stocks led declining stocks by 7 to 5." That is, 7/12 of all stocks went up for the day, and 5/12 went down.

Now, there are thousands of stocks in the market, so 7/12 is most likely just an approximation to the actual number. My question is, how good an approximation is it? You have enough information to answer the question if you make one reasonable assumption.

You can take the division symbol to indicate perfect accuracy, so any errors in the final answer are due entirely to the numbers 7 and 12. So the question becomes how good an approximation are 7 and 12?

-Dan
• Mar 19th 2007, 12:04 AM
CaptainBlack
Quote:

Originally Posted by slobone
The other day I heard the business reporter on the radio say "Advancing stocks led declining stocks by 7 to 5." That is, 7/12 of all stocks went up for the day, and 5/12 went down.

Now, there are thousands of stocks in the market, so 7/12 is most likely just an approximation to the actual number. My question is, how good an approximation is it? You have enough information to answer the question if you make one reasonable assumption.

If there is an infinite population of stocks, and you sample N of which n when
up, and estimate proportion P of the pouulation that went up by p=n/N.
Then if N is large enough n is a random variable with a binomial distribution
with mean

mu = NP

and standard deviation:

sigma = sqrt(NP(1-P))

so p is a RV with SD sigma/N.

If instead we look at the entire population of size N, and find n went up then
there is no error in putting p=n/N as the proportion that went up.

RonL
• Mar 19th 2007, 03:41 PM
slobone
Hmm, not going in the direction I intended so let me clarify a little.

Given a real number (not necessarily rational) between 0 and 1, you want to find the "best" rational approximation to it. You might say for example "half of the voters in the US are Democrats." Now the actual proportion is unlikely to be exactly .50000..., but maybe it's really .49653..., so most people would say 1/2 is the best rational approximation you're going to come up with.

So my original question can be restated, If 7/12 is the best rational approximation to a particular real number, what can you tell me about that number? How good an approximation is 7/12? There's already enough information to solve the problem if you make one reasonable assumption.
• Mar 19th 2007, 04:16 PM
Jhevon
Quote:

Originally Posted by slobone
The other day I heard the business reporter on the radio say "Advancing stocks led declining stocks by 7 to 5." That is, 7/12 of all stocks went up for the day, and 5/12 went down.

Now, there are thousands of stocks in the market, so 7/12 is most likely just an approximation to the actual number. My question is, how good an approximation is it? You have enough information to answer the question if you make one reasonable assumption.

well, it would be hard to tell exactly how accurate this ratio is without actually knowing the figures. How many stocks are they talking about? We don't know. It does, however, seem like they are going by ratios (for every 12 stocks counted, 7 went up and 5 went down), so I don't expect it to be inaccurate enough to cause concern.
• Mar 19th 2007, 05:52 PM
ThePerfectHacker
Quote:

Originally Posted by slobone
Hmm, not going in the direction I intended so let me clarify a little.

Given a real number (not necessarily rational) between 0 and 1, you want to find the "best" rational approximation to it. You might say for example "half of the voters in the US are Democrats." Now the actual proportion is unlikely to be exactly .50000..., but maybe it's really .49653..., so most people would say 1/2 is the best rational approximation you're going to come up with.
.

This is one of my favorited things I studied. The best possible rational approximation* of an irrational number occurs as the continued fraction for the number.

*)We define it the most accurate** rational with a given smallest denominator

**)We define it that |x-p/q|<e where e is the error.
• Mar 19th 2007, 06:45 PM
slobone
Quote:

Originally Posted by ThePerfectHacker
This is one of my favorited things I studied. The best possible rational approximation* of an irrational number occurs as the continued fraction for the number.

*)We define it the most accurate** rational with a given smallest denominator

**)We define it that |x-p/q|<e where e is the error.

That's basically what I had in mind. If p = 7, q = 12, what is the maximum value for e give your *) criterion?
• Mar 19th 2007, 07:05 PM
Quick
There is no way to get an irrational number from the problems you presented.

Let's say that 33 out of 100 random people were guys. That would mean that the percentage of guys is 33.33333333333333333333333333333333333333333333333 333333333333333333333333333333333
or you could just say that it's 2/3 guys.

In this case 2/3 is the exact number, whereas the decimal is not.

A fraction is either just as good or better than a decimal.
• Mar 19th 2007, 07:23 PM
ThePerfectHacker
Quote:

Originally Posted by Quick

A fraction is either just as good or better than a decimal.

I have no idea what you are talking about, but a decimal is a fraction.

Name a fraction which not a decimal and conversly.
• Mar 19th 2007, 11:31 PM
CaptainBlack
Quote:

Originally Posted by ThePerfectHacker
This is one of my favorited things I studied. The best possible rational approximation* of an irrational number occurs as the continued fraction for the number.

*)We define it the most accurate** rational with a given smallest denominator

**)We define it that |x-p/q|<e where e is the error.

To use a colloquialism, isn't this arse about face, the continued fraction gives
the best rational approximation to an irrational number, but here we are
talking about being given a rational approximation to a (not-necessarily
irrational number - in fact it has to be rational as there are a finite number of
stocks) what range of proportions are consistent with that rational approximation.

In these terms we see that there is in fact no answer because we don't know
what psychological factors determined how the number quoted was derived,
i.e. what the rounder/approximater considered an socially acceptable
rounded proportion.

RonL
• Mar 22nd 2007, 10:22 PM
slobone
In everyday life, if somebody uses 7/12 as an approximation, it must be because there is no fraction with a lower denominator that gives a better approximation.

That means the number is closer to 7/12 (=.58333...) than it is to 4/7 (=.57142...) or 3/5 (=.6). So the number must be between .57738... and .59166..., or an error of less than 1% in either direction.