Choosing a Probability Distribution for Visualizing Discrete Data Sets

[Somefantastik]

I have a discrete set of data. I'd like to visualize it probabilistically. Unfortunately, I focused in Num Methods in grad school and am very weak in Probability. Where is a good place to start to visualize this data set using a discrete pdf?

I know a histagram is good to show # of occurrences for each outcome. I also know a cdf plot shows the probability of the outcome being less than some number. But when I start looking at plotting pdf's, there are many functions to choose from and I'm not sure how to go about choosing one, or translating that to a discrete data set rather than a continuous one.

 

[Somefantastik]

mkay so I know that I should make a histogram, normalize the histogram, and then fit a curve to the distribution then.

Now my question is, can I normalize my data before making a histogram, and will that process give me the probabilities on the y-axis?

 

[pmsrw3]

Yes, you can normalize before making a histogram. Suppose, for instance, that you have N measurements, which come as n distinct values x_1, x_2, ..., x_n with frequencies f_1, f_2, ..., f_n. The frequencies are positive integers that add up to N. If you divide each frequency by N, you now have (estimated) probabilities p_i for each x_i that add up to 1. When you make your histogram you'll be binning the x_i, and you get the probability of that bin by adding up all the p_i that go in it. That's the estimated probability of falling into that bin.