Home Product Details Information Sources Gallery

This model has been created to illustrate the profile, properties and applicability of different mathematical distributions.

One of the simplest forms is the Uniform Distribution which has a flat profile. An example of this can be generated by throwing a regular, fair die (dice) and counting the number of times (or 'frequency') that the numbers on each face appear. The expectation is that the frequency of each number will be the same.

In reality after a few throws you may find that one or more numbers seem to occur more often than others. However, as the number of experiments (throws) increases these features will even out.

Fair means the die (dice) is not biased i.e. there is an equal chance of rolling each number. If the dice is biased the long-run outcome (resulting from many throws) will show the frequencies of each number are not even - the bias indicated in the numbers which appear most often.

The model here provides a demonstration of this principle. Once started, the model generates a random number (an integer in the range 0 to 9) and produces a block which is added to the relevant column (0 on the left through to 9 on the right). As the columns grow the model rescales.

In the early stages of the model operation some columns may grow faster than others leading to an uneven pattern. This is a common feature of random sequences where, over a short period, one or more numbers dominate, but as the number of random numbers generated (iterations) increases the profile becomes more even.

The numbers which are generated by computers are termed 'Pseudo-random' because they are not truly random. Computers perform tasks mechanistically which means the sequences of numbers they produce contain repeatable patterns but these commonly only appear after a very large number of iterations.

This simple example is based on random integers - the numbers generated can only take fixed values (0,1,2,3,4,5,6,7,8,9). This type of data is termed 'discrete' because the numbers fit into categories. The greater the number of categories and iterations the smoother the frequency profile. Where numbers can take any real values (e.g. 42.3091) the data is termed 'continuous'.