# Prove the difference

#### Eva B

##### New Member
Dear experts,
I have a product which is made with 3 materials A, B and C, where the composition mix of this product is A – 50%, B – 40%, C – 10%. These add up to be 100% in total.

Then I would like to ask a group of existing customers (e.g. 50 people) on how they estimate the composition mix of the product, that is how much A, B and C in % are used to make the product. so I will get around 50 sets of such customer rated composition mix.

I need help on which statistical test(s) should I use to:
1. prove there’s difference (or no difference) between the customer estimated composition mix against the actual composition mix?
1. Also, how should I prove there is individual difference on customer’s estimated composition mix?

Thank you very much.

#### hlsmith

##### Less is more. Stay pure. Stay poor.
Can a custom respond using any three integers that sum up to 100?

@Miner

#### mahi2pal

##### New Member
It's a multivariate normal distribution. Do multivariate Normal test to see if the Group mean is equal to (50,40,10). to find out the extreme individuals, use the outlier detection concepts in multivariate context.

#### Karabiner

##### TS Contributor
Also, how should I prove there is individual difference on customer’s estimated composition mix?
I am not sure what you mean here. As soon as there is variability in your sample (i.e. not all n=50 give
exactely the same three numbers), it is proven that there are ndividual differences.

With kind regards

Karabiner

#### Miner

##### TS Contributor
This looks a lot like "round robin" inter-laboratory testing. Try searching on the various ASTM or ISO standards for this type of testing to see what statistical analyses are recommended.