Statistical test of experimental data; no expected value?

anonymous

In one of our chemistry class experiments, five groups of students independently came up with the following values for what should be the same constant:

350.9, 300.0, 339.0, 382.0, 362.0

We have no expected value for the constant, only the class data. Is there a statistical test that I can run on this data to determine how "close" or how "near" these data points are? Of course, I could simply take the mean and then determine the percent difference from the mean for each data point, but is there a better way? I have not found anything promising on google, save for the so-called "Pitman nearness test," for which I cannot find a full description.

Does anyone know a statistical test that is relevant to this situation?

anonymous

I would have thought that the standard deviation would do. This is the most commonly used measure of spread and that is essentially what you are talking about.

Has the next answer noted that you don't have a value for the true mean? This is always the case with real research work.

anonymous

Supose you have n (n≥30) values of your experiment:

x₁, x₂, x₃, ..., x₃₀.

You want to make a test about the real mean μ of data.

You can apply the t-test of Student this way:

❶ t = (x* - μ)/(S/√n)

where:

x* = (1/n)Σxi

S = 1/(n-1)Σ(xi - x*)²

S is the population standard deviation.

μ is the population mean.

Suppose the value of population mean is μ = 340, and you have a list of experimental data (n = 30).

Then you will have to test the hypothesis:

H₀: μ = x*

H₁: μ ≠ x*

You can search now some internet sites about this subject.

NOTE: For everything μ value you suppose, you will have to do an hypothesis testing.

Best Regards.