Adjusting for population differences between samples... - (Mar/05/2008 )
Here's the set up:
I have an experiment in which I seeded a specified amount of cells into a dish, each dish being a different condition.
I counted my cells of interest, each sample in triplicate and calculated the average and standard deviation.
Then I plotted these in a simple line chart per condition.
All said and done, they are nice supporting data.
Now my PI is worried because each chart has a different range of cells.
Example:
3 samples under condition A: 1A, 2A and 3A Range of cells: 100-350
3 samples under condition B: 1B, 2B, 3B Range of cells: 280-650
3 samples under condition C: 1C, 2C and 3C Range of cells: 500-800
I am not trying to compare each condition to the other. My purpose is to show the effect on the cells under each of those conditions. Is there a way to calculate or present the data so that ranges are more similar?
I can provide more information but am unsure of what which info would be helpful.
Thanks,
Amelia
I have an experiment in which I seeded a specified amount of cells into a dish, each dish being a different condition.
I counted my cells of interest, each sample in triplicate and calculated the average and standard deviation.
Then I plotted these in a simple line chart per condition.
All said and done, they are nice supporting data.
Now my PI is worried because each chart has a different range of cells.
Example:
3 samples under condition A: 1A, 2A and 3A Range of cells: 100-350
3 samples under condition B: 1B, 2B, 3B Range of cells: 280-650
3 samples under condition C: 1C, 2C and 3C Range of cells: 500-800
I am not trying to compare each condition to the other. My purpose is to show the effect on the cells under each of those conditions. Is there a way to calculate or present the data so that ranges are more similar?
I can provide more information but am unsure of what which info would be helpful.
Thanks,
Amelia
You can perhaps draw a random sample of same size out of the given samples (i.e. you discard some of the samples/data). If you calculate statistics e.g. Anovas and post-hoc tests, then different sample sizes are automatically is considered; generally tests etc have reduced stat. power (slight statistical differences are not considered as significant at a lower level).
Anyhow, if you don't compare the different numbers of different treatments you should not bother imo.
you could also draw a boxplot and see if the results differ by population size.