Hi guys.
I worked out a formula:
λ=f(x)=a+(b-a)/(1+exp(k(x50-x)))
x is the dependent variance while k is uncertain depending on water quality
Now i would like to do data correction on my estimated data (based on the formula above). Is there any recommendation? Can I use probability? The attached graph are the histograms with Δλ and frequency under the different k.
but the Δλ distribution is not normal
It looks like a unimodal, symmetric, fat-tailed distribution. You could try the t-distribution.
As there is a finite and small amount of data, the normal distribution can be used but with caution.
It is suitable to use the (Fisher) F-distribution (the probability density function in your case) for a discrete set of data which will take into consideration the variance based on the discrete amount of data.
You then could attach some "weights" and use your own , modified F-distribution if you want. It is already implemented in Excel, Matlab, Octave and Python.
Franz Nikolaus Pölzl Jochen Wilhelm Sadly, it is not even normally distributed. Is it possible to convert this data to normal distribution? Or I may not use t-distribution or f-distribution. Actually I have tried log, sqrt, arcsin(sqrt) transformation. but none of them works
Non-linear sigmoid models and especially dose-response models can be used easily.
I didn't say that it's normal distributed. I said that - looking at your histograms and assuming that they show an approximation of the density - it might be a t-distribution.
Why don't you upload the data, so one could have a look?
- Badges
- Science topic
- Similar topics
- Correction