# Which statistical test should I use?

Asked by hullcrush (11) October 21st, 2011

I have six bacteria samples, including two controls. I’d like to measure colonies of bacteria in sample vs. control.

I’d like to individually compare samples 1,2,3,4 to control 1.
I’d like to individually compare samples 1,2,3,4 to control 2.
I’d like to compare control 1 to control 2.

The amounts of bacteria will be completely random.

Observing members: 0 Composing members: 0

Cound the colonies and compare. It seems pretty obvious.

XOIIO (18054)

I mean, is there a statistically significant difference between 1 and control etc.

hullcrush (11)

Well, is there? We won’t know until you compare the numbers.

XOIIO (18054)

If you already know that the amounts of bacteria will be completely random, why do a statistical test? :P

dappled_leaves (15780)

we need more information. What are you measuring? What is the difference between the samples and the control? What are your hypotheses? How are you drawing the samples?

wundayatta (58505)

If you want to compare samples 1, 2, 3, and 4 and ask if any of them is different from control 1, then you should do a one-way analysis of variance (ANOVA). If you have a significant difference, you can then do a post hoc t-test (http://en.wikipedia.org/wiki/Student's_t-test) on each comparison (sample 1 vs. control 1, sample 2 vs. control 2, etc.) to see which differences were driving the effect.

You could also do individual t-tests on each (sample 1 vs. control 1, sample 2 vs control 1, etc) but if you do it that way you should correct for multiple comparisons. A pretty straightforward way to do this is using a Bonferroni correction, where you just divide you alpha-level by the number of comparisons you’re doing.

The same answer applies to your second question (comparing samples 1, 2, 3, and 4 against control 2).

Your third situation can be done with either a one-way ANOVA or a t-test; they will yield statistically identical results.

If you don’t have the software you need to do these comparisons send me a message and I will hook you up.

nikipedia (27439)

The software has built in functionality to import a wide variety of data sets and guide you through a battery of available tests.

Academic licensing is at a deep discount.

or

#### Join

This question is in the General Section. Responses must be helpful and on-topic.