Relative Frequencies and distributions i have calculated the relative frequencies for some coefficients and nearly all of them are in the intervall 0,9..1,00.
Now two questions:
a) would it be better to filter the data and just work with the data over a specific thresolds
b) depending on the answer of question a) which would be a possible distribution?
i would like to show the picture but may not yet :-(
But you can try to imagine the way, that i have 10 classes with band with = 0.1 and more than 95% are in the last one from 0.9 to 1.0.
thank you
 A: Comment:
The beta family of distributions has support $(0,1)$ and so is
frequently used to model proportions. Depending on the parameters
chosen, there is a vast variety of possible shapes of distributions
in this family. Roughly speaking, the first shape parameter, often
denoted $\alpha > 0$ governs the shape of the density function near 0;
the larger the parameter the 'flatter' the curve near 0. The second
shape parameter $\beta > 0$ governs the shape of the density function near 1.
The mean of the distribution is $\alpha/(\alpha + \beta).$ You can
read details and see graphs in the Wikipedia article 'beta distribution'.
From what you say, it seems possible that a member of the beta
family with $\alpha$ large and $\beta$ small might fit your data.
Below is the histogram of 10,000 observations simulated according
to $Beta(20, .5),$ along with the density function of that distribution.

Only about 4% of the probability under this density curve lies below 0.9.

Computer code from R statistical software, used for the above:
 x = rbeta(10^4, 20, .5)  # generte fake data
 hist(x, br=20, prob=T, col="wheat")
 curve(dbeta(x, 20, .5), lwd=2, col="blue", add=T)
 pbeta(.9, 20, .5)
 ## 0.04132748  # P(X < .9) = .041

