# Simulation Beta Distribution

Suppose a function such as:

$\phi^{\alpha} (x,y) = \left\{ \begin{array}{l l} \alpha x & \quad \text{if$x < y$}\\ x & \quad \text{if$x = y$}\\ \alpha x + (1 -\alpha) & \quad \text{if$x > y$}\\ \end{array} \right.$

I wish to run a simulation for 5 random walks (each with length 100) on $\phi$ where $x$ is drawn from a beta distribution with parameters (3,1) and $\alpha = 0.5$.

I don't know how to deal with $\phi$ in the random walk process, especially given its 3 conditions.

EDIT: $y$ is the mean of all the previous $x$. For instance, suppose we are at time $t=3$ and observations start at $t=0$. Then, $y_3 = \frac{x_0 + x_1 + x_2}{3}$.

ϕ[x_, y_] := Piecewise[{
{α*x, x < y},
{x, x == y},
{α*x + (1 - α), x > y}}]
ϕ[x, y] • OP notes that $X\sim Beta$ ... which is a continuous random variable. It is not stated what $y$ is, but if $y$ is a constant, or $y$ is an independent random variable, then $P(X=y) = 0$. If so, you can ignore the case $X = y$, because it happens with zero probability. Aug 7, 2014 at 15:31
• Thanks for your comments. $y$ is the mean of all the previous $x$. For instance, suppose we are at time $t=3$ and observations start at $t=0$. Then, $y_3=\frac{x_0+x_1+x_2}{3}$.
– flo
Aug 7, 2014 at 18:05

With the function

f[a_, {x_, y_}] := Piecewise[{{a x, x < y}, {x, x == y}, {1 - a + a x, x > y}}, 0]


you can define a new function, that will perform a single simulation

sim[a_Real, n_Integer] :=
Module[{data = Partition[Riffle[#, Accumulate[#]/Range[n]], 2] &@ RandomVariate[BetaDistribution[3, 1], n]},
f[a, #] & /@ data
]


where the first argument is the value of alpha and the second argument is length of the random walk.
Than 5 random walks of length 100 with alpha = 0.5 can be plotted using

ListPlot[Table[sim[0.5, 100], {n, 5}], PlotRange -> All, Joined -> True, Frame -> True] • Thanks a lot @Karsten! It's starting to make a lot of sense. I just need to work on $y$, since we should observe a convergence towards the mean of x, $E(x) = 0.75$.
– flo
Aug 7, 2014 at 18:51

Here is an approach using MovingMap and f from Karsten 7:

f[a_, {x_, y_}] :=
Piecewise[{{a x, x < y}, {x, x == y}, {1 - a + a x, x > y}}, 0]
s[lst_] := MovingMap[f[0.5, {#[], Mean[#[[1 ;; 3]]]}] &, lst, 3]


Visualizing:

ListPlot[s /@ RandomVariate[BetaDistribution[3, 1], {5, 100}],
Joined -> True, Frame -> True] 