Generating random numbers with known mean but no information about underlying distribution

1 view (last 30 days)
I have some population-level gene expression data, and I'm trying to use it to simulate underlying single cell data. Is there a way in Matlab to create simulated data where I only know the mean and want to make no assumptions about the underlying distributions?
Thanks!
  1 Comment
James Tursa
James Tursa on 4 Mar 2015
Edited: James Tursa on 4 Mar 2015
Do you mean you want to create values from a distribution, and the only thing you know about this distribution is that you have a sample set of data from it? I think you are going to have to make at least some assumptions, like the samples will be uncorrelated. My guess is there may be something in the Statistics Toolbox that can do this for you (i.e., an RNG based on a distribution built from a sampled data set), but I do not know much about this toolbox myself.

Sign in to comment.

Answers (3)

Jan
Jan on 4 Mar 2015
Edited: Jan on 4 Mar 2015
No. To create random data you have to decide for a distribution. Otherwise Matlab cannot guess wehat you might want.
Please explain your intention.

Bryan B
Bryan B on 4 Mar 2015
Fair point. I guess what I don't know is whether or not a given gene may be multimodal or not or what the variance might be so I was trying to go into the analysis as agnostically as possible.

Star Strider
Star Strider on 4 Mar 2015
A very brief PubMed search yields two papers that might shed light on the distribution you need:
Those were the only ones whose abstracts seemed relevant. I obviously did not look at all 15276 papers that search produced. It seems to be a problem that defies easy assumptions, but is an area of very active research.

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!