# Measuring variation - (Jan/23/2014 )

Hey friends,

If I measure one parameter by different methods (A,B,C, D, E) and I get different readings how Can I explain variation statistically?

Depends on the variable - discrete? continuous?

If you use different indices on the same data you might get differnet results, because one index might be more sensitive to cetain aspects of the data, if that was your question. It basically depends on the mathematical formula of each index.

For example, in community ecology there are many indices of species diversity. Some are more sensitive to species that appear at low frequency, so the difference in diversity between two communities or samples might be more significant with one index than another if one sample has many species at a low frequency.

Again, I'm not sure if I understood your question correctly.

Thanks for reply.

Here's other version of same question. If I measure blood glucose by five different methods (A,B,C, D, E) and gel results 82, 84, 86, 89, 93 mg/dl resp. how can I explain variation produced by different methods. (if I am not aware of normal value!!!)

@ Bob:- Discrete variable.

If you have only one measurement per method you can do not much with statistics as you don't know if it's variation by chance (random effect) or by method.

You should do a series of measurements and then compare means.

thanks.

Wouldn't doing a series of measurements and then comparing means is same as comparing individual readings? if individual method don't show variation.

Inbox on Fri Jan 24 21:11:33 2014 said:

Wouldn't doing a series of measurements and then comparing means is same as comparing individual readings? if individual method don't show variation.

But exactly to prove that each individual method does not show a variation, you need several measurements and take the mean. For example, the method giving you a value of 82 seems to give a lower value than the one with 89. But if you do several experiments with the first method and find that the values you get vary between, say, 81 and 90, and your second method gives values between 83 and 91, the difference might not be statistically significant.