which test to use for comparing samples with single observation???


New Member
I am testing percent genotoxicity of water samples using a bioassay which is expensive to perform and therefore cannot have replicates. Each sample is analyzed only once. Same goes for negative control. Now i want to compare the genotoxicity of all samples with each other and with negative control. How to do that as ANOVA requires at least triplicate values for all samples. So which other test is possible. Please help!
Thanks in advance!!!


New Member
Show us an example with real values of say, 10 values, 5 treatment and 5 "negative control".

Anova (analysis of variance) does NOT require that!
Thanks for replying Greta Garbo

Suppose I have only one negative control with genotoxicity percentage 2% and three samples a, b and c with genotoxicity percentage 6%, 7% and 5% respectively in season 1. Then I have second season samples a,b and C sampled from same location as season 1 with % genotoxicity 5%, 8% and 4% respectively. Negative control for second season is again 2% genotoxic. These all are single observations. No replicates. I want to compare each sample genotoxicity with respective seasons negative control and samples with each other within a season and even the seasons to each other. What to do. You said that ANOVA is possible. Pls explain how! If I would have had triplicates, I would have gone 2 way ANOVA to compare both samples and seasons simultaneously as I would have used each sample as group with three observations. But what to do in this case??? Pls reply It would be great help!!!
The word 'sample' is used differently in statistics as compared to among the chemists.

If you have 20 buckets of water from 20 lakes, one from each lake, then you would have a sample of 20. (We statisticians say that you have one sample.) Then the "unit of investigation" would be the bucket of water. Of course there would be a random variation between the buckets.

Maybe you have measured each bucket 3 times, so that you get 3 values for each bucket. But the sample size would still be 20 (n=20). Of course there there can be a random measurement error so that there is a variation among the measurement. There are two sources of variation, one between the buckets and one between the measurements. (By the way, it better to have 60 buckets of water (from 60 lakes) and one measurement per bucket than to have 20 buckets and 3 measurements per bucket.)

What is what among the "a, b and c" values and negative control. Please explain more.
It seems like we have lost the original poster.

Sometimes it happen that you get interested in how something work, ask questions, but then the original poster disappear.