I´m doing a research testing the viability of p gingivalis (ATCC) against ozone and I need to know if it is necessary to do a sample size calculation.
These two references shows that you can use n=3 to 6. It is acceptable for microbiology experiments.
Article Effect of salivary pellicle on antibacterial activity of nov...
Article Mechanically Responsive Antibacterial Plasma Polymer Coating...
To better understand your question, what do you call "sample size»? Do you mean the number of bacteria before treatment? You have to do an estimation of the number of bacteria, using cultural and/or spectrometric technics.
Thanks Laurent for your answer. I not mean the bacteria number (i know it should be between 25 - 300 cfu for statistical reason). I mean is it necessary to calculate a sample number of experiments to find differences? (how many times and in how many plates do i need to apply ozone), or just duplicate or triplicate. The thing is i can't find any reference that tell me "yo don't need it, it is an vitro study and you need to do the experiments in duplicate or triplicate"
I´m comparing the viability of p gingivalis against ozone and the negative control is the exposure of the same atcc p gingivalis to oxygen (during the same time of course) to evaluate the antibacterial effect of ozone.
Dear Franz,
It depends on the assay you are aiming to accomplish. If it is a colorimetric assay, agar plate diffusion or counting of bacteria, normally triplicate is acceptable for each group. If you are working on microscopy quantitative measurements, you must have a larger number of images of each group, 5 or 10 images (for 3D and 2D images, respectivelly) of each group os samples, and then triplicate the experiment for statistical reasons.
Regards,
LP
These two references shows that you can use n=3 to 6. It is acceptable for microbiology experiments.
Article Effect of salivary pellicle on antibacterial activity of nov...
Article Mechanically Responsive Antibacterial Plasma Polymer Coating...
I agree with Lilan Response.
Alternatively you can base your number of replicate on the statistical power of your experiment, implying that an important difference needs less replicate than a small disparity. Usually triplicate will be ok, but for exemple we have used 5 mice in our paper on ozone effect against toxoplasma :
(http://www.sciencedirect.com/science/article/pii/S0304401708000903)
Good luck with your experiments.
If your response is categorical on a binomial basis (i.e. viability/no viability), you can use a binomial test to prove if your event is aleatory or not, then n=5 is fine (if all the experiment's responses are "no viability", the p-value is 0.0312).
But if your variable is quantitative (cfu) after the ozone treatment, you must calculate the number of experiments based on the statistical power of the test that you will perform, and it depends too to the number of treatments that you will prove (different ozone concentrations and/or control groups).
Another way also acceptable for quantitative responses, is to perform a regression test, and for that 5 to 6 ozone concentrations with triplicates (n=3) will be OK. Then you will estimate with a confidence interval the effect or the ozone concentration in the bacteria growth or vice versa (estimating a ED50).
Thanks Federico, very interesting answer, do you have a reference of that? thanks in advance
Franz, you can find the foundations of binomial hypothesis test in chapter 13 of this book (in Spanish):
Milton, J. (2007). Estadística para la Biología y Ciencias de la Salud. Tercera Edición ampliada. Editorial McGraw Hill.
You can use this on-line calculator:
https://src.alionscience.com/cgi-src/calc.pl?pval=0.5&nval=&rval=0&clval=95
Type: p=0.5, r (fails) = 0, CL = 95 then click on CALCULATE and the “n” will be 5. This is restricted to the hypothesis that your event is aleatory (Ho: p=0.5) or not and you espects no fails (all your assays must be “inhibition”).
For the regression analysis of quantitative responses in a dose-response assay, the US Pharmacopeia could be helpful (because it is based on ATCC bacteria), that recommends five levels and triplicates for a standard curve of antibiotic potency assay (that could be used in other assays, including assessment of lineality, because statistically you will have a good estimate of the standard error of the best fitting curve). If you are sure that your dose-response curve is linear you can reduce your points (3 or 4); but if the curve is exponential you will need at least 5 points (or maybe more). Remember that a negative control group will be necessary to assess the optimal growth of the bacteria (without ozone), and then your dose-response curve could be constructed using the responses on terms of % of viability or % of inhibition, in order to estimate de ED50.
I hope that this could be helpful for you.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size