Author Topic: correlation coefficient  (Read 1858 times)

0 Members and 1 Guest are viewing this topic.

Offline waru

  • LV1 Newcomer (Next: 20)
  • *
  • Posts: 6
  • Rating: +0/-0
    • View Profile
correlation coefficient
« on: September 09, 2012, 12:14:08 am »
hi everyone  i came here to see if i can get some help from you this is something that is killing me because i can't find it...

i have a TI nspire cas and need to find the correlation coefficient of  graphic  (2-sin(2x)-e^x) i have no idea on how to do it so plz i beg your help


thanks a lot

Offline leafy

  • CoT Emeritus
  • LV10 31337 u53r (Next: 2000)
  • *
  • Posts: 1554
  • Rating: +475/-97
  • Seizon senryakuuuu!
    • View Profile
    • keff.me
Re: correlation coefficient
« Reply #1 on: September 09, 2012, 12:15:22 am »
I thought the correlation coefficient is based on a series of points from which a best-fit line or curve is derived, indicating how closely the line fits? You'd have to provide the points from which that graph is derived.
In-progress: Graviter (...)

Offline waru

  • LV1 Newcomer (Next: 20)
  • *
  • Posts: 6
  • Rating: +0/-0
    • View Profile
Re: correlation coefficient
« Reply #2 on: September 09, 2012, 12:23:11 am »
and can you tell me please how can i do that? i really have no idea on how to do anything with graphics  whit the TI  sorry

Offline Deep Toaster

  • So much to do, so much time, so little motivation
  • Administrator
  • LV13 Extreme Addict (Next: 9001)
  • *************
  • Posts: 8206
  • Rating: +758/-15
    • View Profile
    • ClrHome
Re: correlation coefficient
« Reply #3 on: September 09, 2012, 12:29:20 am »
The correlation coefficient is a property of a set of data points, so we need to know the data that the question asks about.

Can you post the question that asks you for the correlation coefficient?




Offline waru

  • LV1 Newcomer (Next: 20)
  • *
  • Posts: 6
  • Rating: +0/-0
    • View Profile
Re: correlation coefficient
« Reply #4 on: September 09, 2012, 12:31:35 am »
ok basically it gives me the function (2-sin(2x)-e^x) and asks me to find the "r" value of the graphic

Offline ruler501

  • Meep
  • LV11 Super Veteran (Next: 3000)
  • ***********
  • Posts: 2475
  • Rating: +66/-9
  • Crazy Programmer
    • View Profile
Re: correlation coefficient
« Reply #5 on: September 09, 2012, 12:58:05 am »
waru if it doesnt give you a set of points with that you cant find it. unless it means something else by "r" value
I currently don't do much, but I am a developer for a game you should totally try out called AssaultCube Reloaded download here https://assaultcuber.codeplex.com/
-----BEGIN GEEK CODE BLOCK-----
Version: 3.1
GCM/CS/M/S d- s++: a---- C++ UL++ P+ L++ E---- W++ N o? K- w-- o? !M V?
PS+ PE+ Y+ PGP++ t 5? X R tv-- b+++ DI+ D+ G++ e- h! !r y

Offline waru

  • LV1 Newcomer (Next: 20)
  • *
  • Posts: 6
  • Rating: +0/-0
    • View Profile
Re: correlation coefficient
« Reply #6 on: September 09, 2012, 04:18:47 pm »
uhmmmmm thats the thing it seems that they want me to get points out if that function and then get the correlation coefficient

Offline Deep Toaster

  • So much to do, so much time, so little motivation
  • Administrator
  • LV13 Extreme Addict (Next: 9001)
  • *************
  • Posts: 8206
  • Rating: +758/-15
    • View Profile
    • ClrHome
Re: correlation coefficient
« Reply #7 on: September 09, 2012, 04:22:38 pm »
uhmmmmm thats the thing it seems that they want me to get points out if that function and then get the correlation coefficient
That won't work, because the correlation coefficient represents how well an equation matches a set of data values (such as from a survey or experiment). If you try to calculate the correlation coefficient of that equation with itself—well, of course it's a perfect match :P It's like playing Hangman with yourself.

I have the feeling that's not the entire question. Where did you get this question, and is there a data table somewhere near?




Offline waru

  • LV1 Newcomer (Next: 20)
  • *
  • Posts: 6
  • Rating: +0/-0
    • View Profile
Re: correlation coefficient
« Reply #8 on: September 09, 2012, 04:25:49 pm »
my professor gave it to me he just gave some random functions and said " graphicate them then get the correlation  coefficient out of them " and no there is no table near as you said i need points but he never gave us any

Offline AngelFish

  • Is this my custom title?
  • Administrator
  • LV12 Extreme Poster (Next: 5000)
  • ************
  • Posts: 3242
  • Rating: +270/-27
  • I'm a Fishbot
    • View Profile
Re: correlation coefficient
« Reply #9 on: September 09, 2012, 04:42:37 pm »
Technically you CAN generalize correlation coefficients to infinite sets, but the set typically has to remain bounded and there needs to be some reference curve with which to compute the coefficient. The curve (2-sin(2x)-e^x) is NOT bounded below and no reference curve is provided. Even if you assume a linear curve, the fact that the function given approaches negative infinity faster than a line means the result isn't in the reals (in other words, it's meaningless).

Unless you can provide an interval and spacing between points, the question isn't really answerable.
∂²Ψ    -(2m(V(x)-E)Ψ
---  = -------------
∂x²        ℏ²Ψ

Offline waru

  • LV1 Newcomer (Next: 20)
  • *
  • Posts: 6
  • Rating: +0/-0
    • View Profile
Re: correlation coefficient
« Reply #10 on: September 09, 2012, 06:06:41 pm »
maybe the professor didn't explain it well here is the thing i need to do a numerical analisys  of functions the asked me to use some methods (secant, flase position,newton ,etc) so i need a value out of the graphic ( i think it is the one i mentioned ) BUT it may be be the slope  i have that problem please help me on this because its killing me

Offline AngelFish

  • Is this my custom title?
  • Administrator
  • LV12 Extreme Poster (Next: 5000)
  • ************
  • Posts: 3242
  • Rating: +270/-27
  • I'm a Fishbot
    • View Profile
Re: correlation coefficient
« Reply #11 on: September 09, 2012, 09:10:14 pm »
All of those methods are typically used for what is called root finding. Since you mentioned numerical analysis, in which root finding is a very common operation, I'll assume that's actually what they meant. However, this is probably also homework, so I won't give you a complete answer either. You'll have to figure the rest out on your own.

Let's begin:

Rule 1) In numerical analysis, always plot your function first. This will tell you something about its behavior and how best to approach it. It will also help alert you to any potential funkiness with how numerical methods handle the function. Here's the function you mentioned, (2-sin(2x)-ex) between x=-5 and x=2.2



Of course, that doesn't really tell us anything about the behavior of the function outside, so let's look at the contributions of the terms 2-sin(2x) and -ex to the function...



It's a lot more clear now, isn't it? As x→-∞, the function oscillates between 1 and 3, with a global maximum approaching 3. This is because the -ex term falls to 0 when x→-∞.

As x→∞, of course, the behavior changes radically. The sine term continues oscillating, but it's quickly overtaken by the negative ex which grows, well, exponentially. As you would expect, the value of the function approaches ∞. Take careful note of the term approaches. At no point on the real number line is the function ever "equal to" infinity in any sense. Infinity is not a real number and we can't use it as if it were one.

Now that we know how the function behaves, we can start with the more in-depth analysis. As you mentioned, we're going to be using root finding methods, so that means we'll be looking for roots. But of course, you need to prove that there are roots in the first place. One might think that the Fundamental Theorem of Algebra would solve this problem for us. One would also be wrong. Take careful note of the conditions of validity for that theorem: It only applies to non-constant polynomials. Neither the Sine function nor the exponential function are polynomials, so it doesn't apply here. However, we CAN still prove that there's a root, namely with the Intermediate Value Theorem. As before, we need to check the assumptions. The only assumptions we need to be concerned with are those for continuity and smoothness, both of which are easily determined by the fact that sine and exponentials, as well as all finite functions composed by elementary operations thereof, are analytic. Thus the function we have is analytic, and hence, both continuous and smooth. Accordingly, we can use the IVT. Taking this, we know that the function has a maximum at 3 and a minimum that approaches -∞ (to be perfectly rigorous, there actually is no minimum of this function, but that's not really relevant here). Therefore, there exists a root of the function in the reals.

Now that we've done all the ground work, we can actually go ahead and try to find that root. At this point, I'm kind of wanting to get back to my movie and google has perfectly good explanations of how these work. Just suffice to say that the end results will be pretty much the same (~x=0.33) if you choose the proper starting points.

Secant Method:
http://en.wikipedia.org/wiki/Secant_method
http://mathworld.wolfram.com/SecantMethod.html

Method of False position:
http://en.wikipedia.org/wiki/False_position_method#Numerical_analysis
http://mathworld.wolfram.com/MethodofFalsePosition.html

Newton's Method:
http://en.wikipedia.org/wiki/Householder%27s_method
http://en.wikipedia.org/wiki/Newton%27s_method
http://mathworld.wolfram.com/NewtonsMethod.html
∂²Ψ    -(2m(V(x)-E)Ψ
---  = -------------
∂x²        ℏ²Ψ