Effect of varying number of bars | Vose Software

Effect of varying number of bars

The number of classes used in a histogram plot will almost always affect the relative frequency (but not probability density) scale of the y-axis. Clearly, the wider the bar width, the more chance there will be that values will fall within it. So, for example, by doubling the number of histogram bars, the relative frequency scale will approximately halve.

In plotting a histogram, the number of bars should be chosen to balance between a lack of detail (too few bars, on the left) and overwhelming random noise (too many bars, on the right).

When the result of a risk analysis model is a discrete distribution, you will need to experiment with the number of histogram bars to get the best plot. If possible set the number of bars to be some multiple of the number of allowed values in your distribution; as this will prevent interference patterns:

See Also

 

Navigation