Dear community members,
I performed an analysis of customers inter-arrival times for my case and I have a list of 196 observations (each observation is the inter-arrival time between 2 customers). I used the data fit tool but no distribution fits the data. Then I’m trying to use an empirical continuous distribution.
The problem is that I should define the number of classes, the upper bound and the lower bound of each class.
I thought I could use as classes the same that the data fit creates (picture below)
It actually tells me that I could use 14 classes, but what about the lower and upper bound of each class?
Do you have any suggestion on that?
Solved! Go to Solution.
I'd recommend to choose for the boundaries the mid value of the intervals:
0, (210+350)/2, ~56
(210+350)/2, (350+490)/2, ~31
thanks for your reply.
I just saw that in the tab "evaluation" there is "frequency table". This is the view
I guess it is the way the Datafit divided the input data, along with the the frequencies. Is this good for the empirical distribution? I mean, can I just insert these data instead of doing the calculation you suggested before?
Right, this is the better choice, since you can just copy these values.
I'm facing a similar situation, I have different samples of failures times, I used the Data Fit Tool to find a data distribution, but many of my samples don't feet to any, so I need to use several empirical distributions. Here is the question.
¿Has a simulation model reliability when it uses several empirical distributions?
I think the problem is in my data samples because even analyzing and doing an exercise where I eliminate the outliers, the sample doesn't feet to any distribution.