Please help me set up statistics for my experiment?

Question

i tested the effects of plants on bacteria. i don't know how to get the statistics. please provide some advice for me. i used four different plants and tested them on three different bacteria. the number is the number of millimeters moved. here's one set of data (with three runs/trials of each)

serratia

grapes seed extract 0mm, 0mm, 2mm

ginger 0mm, 0mm, 7mm

holy basil 4mm, 2mm, 3mm

mustard seed 0mm, 0mm, 0mm

erythromycin 5mm, 7mm, 10mm

amoxicillan 3mm, 2mm, 2mm

i have no idea how to do stats! please help me! thank you!

Answer 1

I want to help, but I'm unclear on at least one thing: you mention using four different plants on three bacteria, but you list "serratia" without any data, and data for six entries (four plant names, and what appear to be two antibiotics). Can you clarify what exactly you did?

If I understand your methods correctly, you're a bit limited in terms of a formal statistical test. You have four treatment groups, and only three replicates of each group. Normally, you would run a one-way ANOVA, and then apply post hoc tests (e.g., Tukey test) to test which groups are different from one another if you find an overall difference. In your case, without running the test*, I would be surprised if you will find differences, given the (1) low replication, and (2) large within-group variance. What I mean by the latter: if you look at the ginger treatment, twice the bacteria didn't grow/move at all, and once they grew/moved 7 mm. So, that reduces your confidence in the treatment mean, and will make those error bars overlap with those from the other groups.

*For fun, I did run it (ignoring the fact that normality of variances is almost certainly violated), and the overall P value was ~ 0.35. So, no difference among the treatment groups overall.

For the antibiotic case, you could try running a simple two-tailed t-test. Doing that with an online calculator, I obtain a not-quite significant result (P = 0.079), if I use the Welch t-test, which accounts for non-equal variances between the two groups. Of course, with n = 3 in each group, you really should run a nonparametric test, but I think you can be excused here.

For the first part, visit: http://faculty.vassar.edu/lowry/ank4.htm...

For the second, visit:

http://www.graphpad.com/quickcalcs/ttest...