Calculating the "stillness" of numerical dataset I am a software developer and not very proficient in math. There is a dataset of type [0.84980994, 0.117350034, 47.58483789, ....] 
I want to calculate some evaluation of a rate of change of $X_n$ and $X_{n-1}$ of the list. What kind of evaluation can be used?
EDIT: Maybe this is not clear from the original question, but I want to compare the evaluation of the dataset with an empirically derived value to test a hypothesis. 
My hypothesis is that the level of change of the dataset (differences between the values) is less than a specified threshold. 
Should also note that the difference matters only between neighboring samples, not the whole set of values.
 A: You seem a little vague as to the criterion for stability. 
Looking into the literature of a couple of methods that have been
successfully used may help you crystallize what you mean by stability.
Control charts, long and widely used in the field of quality management,
typically consider a variety of indicators that a process may have
gone 'out of control'. [Roughly, a few criteria are: too many observations in a row above
(or below) the mean; too many consecutively increasing (or decreasing)
observations; going beyond limits such as $\bar X \pm S$, where
the mean $\bar X$ and the SD $S$ may be fixed (based on prior data)
or updated at each new observation.]
Also, there are a number of criteria for stability vs. change from
past behavior in the field of time series, especially as used by
economists. You might start by looking at 'autocorrelation' for
various 'lags'. The autocorrelation function (ACF) is also used by statisticians
using Markov Chain Monte Carlo methods to judge when or whether a
simulated Markovian progress reaches 'steady state'.
Also there is the statistical ideas of outliers, often as 'detected' by
boxplots or (especially among regresion residuals) by Studentized ranges
of excessive absolute value.
