Calculate Prediction Accuracy Using Statistical Test

[Yann]

I have two sets of data. The first is the observed value for a number of objects, the second is the predicted value for those objects. I want to know how much, in %, my equation is able to predict. What's the name of the statistical test to know that ?

 

[HallsofIvy]

I have two sets of data. The first is the observed value for a number of objects, the second is the predicted value for those objects. I want to know how much, in %, my equation is able to predict. What's the name of the statistical test to know that ?

What "equation" are you talking about? Are you under the impression that this question makes any sense at all?

 

[Yann]

Lets say you have a number of species and you know their size S;

(S1, S2, S3 .. Sn)...

Then, you have an equation to predict the size on those animals. So you have another "vector" with predicted size P;

(P1, P2, P3 ... Pn)...

In an article, the author said his equation (the one he used to get P1, P2...) explains X% of the variation in size, but he doesn't give any clue to the method he used to find that %. I just want to know what's the name of the method he used, how can you know how much the predicted size explains the variations in the observed size ?

 

[cactus1003]

What I understood it correctly is as follows:

U have an equation y=f(x), to find the size of the animal.
Now to find its accuracy, u take a sample of size n (i.e. ur observed value) n try to find the % of variation by comparing the values of actual n observed one…
One simple method is
Find out which distribution the random variable X (the size) follows.
If it follows normal distribution, then try to estimate the mean n variance of it by using some unbiased estimator (for any other distribution, estimate the unknown parameter of the PDF).
Hope u can do the needful… good luck