1. ## Analyzing statistics data!

Hi guys, first off, I'm new to this forum... hi everybody!

Lately, I've been working with data for comparing peptide activity on modified peptides. We have a parent peptide (full activity), and a control peptide (minimal activity).

Between them are a range of different activities. We have means, n values and SEMs for all the different values.

The data looks sort of like this (the numbers are arbitrarily assigned):

Parent: value-100, SEM-10, n=30
Control: value-20, SEM-5, n=25
A: value-40, SEM-6, n=27
B: value-70...
C: ...
D: ...
etc.

The data is assumed to be Gaussian distribution.

My question is this: Is there a statistical test that can help me determine whether the experimental peptides (A, B, C... etc.) are EITHER significantly different Parent OR Control?

The reason I ask this question is because we've tried both unpaired T-tests and ANOVA analysis with a Dunnett's post test (along with a host of other 1-tailed ANOVA tests + post test) . However, in every situation, we've found that certain peptides (peptide D for example) is NOT significantly different from the control OR parent peptide.

We've also found that certain peptides (peptide E for example) are significantly different from the control AND the parent peptide. How is this possible and can it be resolved?

We would ultimately like to be able to say that "Peptide X is significantly different from (Control OR experimental)", but with the statistical data we currently have, this is not possible.

Thank you guys in advance for any help and I apologize for the long post!
-Elton

2. ## Re: Major help with analyzing statistics data!

i dont have a test for you, but consider these:

1) Look up the required assumptions for the tests you tried and check that they are satisfied by your data. You may find that your data dont meet the requirements for some of the tests, leading to contradictory results.

2) Remember that there is a probability of getting the wrong result in a hypothesis test and so contradictory results should be expected from time to time.

3) if it really is the case that each of your samples will be either "parent" or "control" you should be able to get a probability of each with a bayesian approach.

3. ## Re: Major help with analyzing statistics data!

Thanks for the advice! I've checked the assumptions and it we seem to fulfill them. As for your third point, how would I go about doing this approach? I did a quick search but was only able to find information about probability. Is there a a statistical test somewhere?