# Bimodal distribution analysis

#### WonderingBeta

##### New Member
Hello everyone,

I need to define the loss of some cells with differential size. In the control population, size distribution is bimodal, each peak seems pretty much gaussian except for the overlapping tails (see attachment). In my cases, cell loss is obvious even though with different magnitude between subjects (from 30% total remaining cells to less than 5%), but I wanted to explore if the loss is homogeneous between the two cell families or if one is more affected.
I came up with two ideas till now:
- categorize the variable in intervals, defining the two peak interval in control population. Then, calculate peak ratio in controls and cases and t-test between the two groups
- pool the control data, use a steam-and-leaf plot on the continuous variable to identify the two peak value, then analyze only the outer tails (the left one for the lower values and the right one for higher values) to get each curve standard deviations. Select the data to include in each subgroup using a one or two-standard deviation cut-off from each peak, then testing that ratio

Am I completely wrong? better ideas?

#### Attachments

• 16.9 KB Views: 2

#### katxt

##### Active Member
I need to define the loss of some cells with differential size.
This problem sounds interesting, but I don't understand the setting. Can you explain more clearly please.

#### WonderingBeta

##### New Member
This problem sounds interesting, but I don't understand the setting. Can you explain more clearly please.
I'm analyzing nerve fibre loss in a patholoci cohort and comparing it with a normal control population. For each case I have the area of every nerve fibre in a certain nerve. Two family of fibers of different size form the great majority of the normal nerve, resulting in a bimodal distribution when nerve fibre size is plotted against numerosity (see previous table). Also, I have no third means apart from size to classify each nerve fiber.
In my case, I need to check if the condition results in a similar loss of both families of fibers, or if one is more affected than the other.
Let me know if I need to better rephrase

#### katxt

##### Active Member
Let's see if I have this clear.
Your data is in two parts - the distribution of healthy fibres in your diagram, and another similar set of diseased fibres.
The question is - is the diseased distribution significantly different from the healthy distribution?
Is that more or less the situation? kat

#### WonderingBeta

##### New Member
Let's see if I have this clear.
Your data is in two parts - the distribution of healthy fibres in your diagram, and another similar set of diseased fibres.
The question is - is the diseased distribution significantly different from the healthy distribution?
Is that more or less the situation? kat
Yes, with the important point being not if there are less overall fibers, but if the two peaks behave differently

#### katxt

##### Active Member
One problem that may come up is independence. Do thicks and thins occur in patches? If you find a thick one is the one next to it more likely to be also thick rather than thin? or is there no connection?

#### katxt

##### Active Member
Are you still looking for ideas? kat