Dear Kamilla,

You always need to be careful when you are making a qualitative assessment of data presented in a graph and in comparing this to the results of an appropriate statistical test.

I normally plot the means with standard errors for my graph, but this is not appropriate if you want to compare visually two mean values on the same graph. If you plot the 95% Confidence intervals instead of the standard errors, then assuming your data are Normal etc, then means with over-lapping 95% CI's are not significantly different (p < 0.05).

I often use an ANOVA / post-hoc test such as Turkey-HSD to see which means are different from the others. Again, I plot the means with the standard errors, but I then add the 'letters' from the Turkey-HSD test onto each of the means - if means are not connected by the same letter, then they are significantly different (alpha = 0.05).

Your two-way ANOVA will let you look at interactions as well as differences between means for your two treatments. At the end of the day, you need to trust your statistics - and not your eyes – as we have a terrible habit of believing what we want to see. Statistics is there to provide an unbiased view of our results - so long as the statistical tests have been done correctly.

Andrew

Hi

A method for analysis of real world data, called the Persian curve, devised. It is a generic method which is suitable for analysis of natural phenomena, including your data. A file containing an overview of the Persian curve attached. It will answer your question. I am ready to answer any question arise.

best regards

Greetings, Kamilla

I believe the reason is that the type of data that you have, OD data of bacterial culture, will be mostly less than 1, and all the values will be close to each other, especially when the standard errors between replicas are relatively high, so I recommend that you use plate count instead of OD and calculate log reduction, not OD difference.