# gene total volume test

1, I want to compare the FGF gene total expression level (sum level from all the cells) in two groups of cells. Suppose group A has 500 cells and group B has 100 cells. 2, gene FGF level in the two groups has no difference in each cell. 3, sum each cell's gene FGF level should have a difference in the two groups. Question: how to test the total volume statistical significance？ My idea is set randomly split into three or more groups(in the figure)

I want to know if this treatment is reasonable.

• Can you explain what you mean by "fold"? Can you also explain why you need to split each group into "3 or more" groups? Or for that matter any subgroups? Also, why is assumption 2) needed? Also, why look at the sum total volume, when the sample sizes are not the same? Instead of looking at the mean expression level, and just compare the mean of the 2 groups (1 with 500 cells, the other with 100 cells)? With a good old 2 sample Welch t-test? You should clarify your problem statement... Commented Jun 4 at 23:59
• See earlier comment Commented Jun 5 at 0:00
• @jginestet Can you explain what you mean by "fold"?' fold= (mean gene level in group A) /(mean gene level in group B). 'Can you also explain why you need to split each group into "3 or more" groups?' to get more means to calculate a p-value, they are randomly split. Commented Jun 5 at 15:45
• @jginestet • 'Also, why is assumption 2) needed? Also, why look at the sum total volume, when the sample sizes are not the same? Instead of looking at the mean expression level, and just compare the mean of the 2 groups (1 with 500 cells, the other with 100 cells)?' It is what I observed in a single cell datasets. More cells express a gene means more impacts in tissues although the gene expressed same in each cell. "With a good old 2 sample Welch t-test? " yes, so I split the big samples into smaller to get more than three sum values to compare. Is it reasonable? Commented Jun 5 at 15:46
• Can you measure the gene expression of each cell individually, or really only collectively, for a group of cells? I assume only collectively, but... If you can measure individually, then just do so, and compare the means with a 2-sample Welch t-test. If you can't, then yes, you need to create several subgroups; but you idea to have 3 subgroups for each cell type only gives you a sample size of 3. Much too small to have any power (unless the difference is enormous, but then, who needs statistics?). A minimum of 12-20 (per cell type) would be more reasonable. ...cont Commented Jun 5 at 20:25