Statistic help required?

[Blandmum]

any Statisticians on MN?

By brother is also a biology teacher and has been asked to help with some A level C/W statistics.

The experimental design is this

The effect of two different antibiotic drugs at 5 different dilutions each.

Should the student use a Spearmans rank for each drug?

Should they use T test between the different drugs?

Help????

I would think (very rusty at this, but I might be right ) t-test only if the sample sizes are big enough (at least 30 observations for each dilution, off top of my head), and independent observations each reading.

Presumably sample sizes are much smaller so a non-parametric test is needed. This page discusses non-p equivalents for t-tests.

so a wilcoxon, do you think?

What is actually being measured? i.e is the the data categorical or a continuous variable? If a continous variable and normally distributed an ANOVA? COuld be completely wrong of course - its been a while ......

It think that it is catagorical, because there are two different, unrelated antibiotics, with 5 dilutions of each. there is no interaction between them, because they are never tested together, which I thought would exclude a two way ANOVA

How is "the effect" measured -- what kind of units? Categorical (eg., it worked or it didn't), ordered (how well it worked to kill bugs, on a scale from 1 to 5), or ratio (eg., % of bugs killed in a petri dish in a given time frame). I was assuming ratio, but too small a sample per dilution to do any parametric testing. Plus, variances in each sub-sample have to be equal for ANOVA.

I think halster was closer to the mark than my first thinking... I think your bro wants Kruskal Wallis test (a non-parametric equivalent to ANOVA). You need ordered or ratio data for that (see assumptions as described in link).