Likelihood: Analysis After Probability?

Question

I am struggling with the analysis of my data. I am interested in hearing suggestions about how to improve that analysis.

I am performing an "Event Probability" study. I know the general "probability" of all possible events presented by my data. I have further scrutinized that data and calculated the probability of those same events given the occurrence of another event--a standard "conditional probability" analysis. After performing the "conditional probability" analysis, it is very clear which event is the most probable (let's call it the "Majority Event"). Typically, the "Majority Event" will have a probability of 35% whereas the remainder of events (let's call them "Minority Events") may have probability values of 15%, 13% etc.--which, in the end, is a hefty 65% of all other events in the aggregate.

Is there a way now, using some sort of "frequency analysis" formula or "distribution analysis" formula to determine "Likelihood" over and above probability.

What I mean is: The most probable event, of course, is not always the actual event that occurs. If there are 8 possible outcomes, and one has a probability of 30%, and all the remaining events have a probability of 10%, the 7 less probable events, in the aggregate, make up a much larger percentage of events than the most probable event. How, then, do you take this probability analysis to a higher level and determine whether the most probable event will occur, or one of the less frequent events that, as a whole, represent a greater percentage of total events.

Answer 1

In situations like this, I have found that the most useful technique for further analysis is using something called the Monte Carlo method. This is where you write a simple computer program to run thousands of events with your given probabilities. Then the results will tell you what your ultimate expectations should be on a long-term basis. You can then use these results to verify your calculated correlation coefficients and conditional probabilities.

There are two advantages of using a Monte Carlo technique:
1. It eliminates the need for what can become complicated analysis that required deeper mathematical knowledge
2. The results can not be disputed (as long as the algorithm is verified to be working correctly). I have found that computational results are often disputed by people who do not understand math and can not accept any non-intuitive results.

Frankly, your description and additional commentary sounds very heady but does not explain how the "likelihood" you are seeking is any different from the probabilities" that you have already determined.

Answer 2

So far as I can tell, you aren't doing anything especially difficult, but your terminology is really screwed up. Go find a real live Stats major or social science grad student who can help you sort your problem out. I think you're talking about 2 events, where the second one has 8 outcomes, for instance, but you're conflating a whole bunch of terms.

Answer 3

you are able to determine those with some actual numbers. a million/(3 + 4) = a million/7. yet a million/3 + a million/4 = 7 / 12, that is a few distance greater. So the considerable one can't be precise. the 2d is fake, as you regulate into conscious of; 3(a + b) is 3a + 3b.

Answer 4

yaho answers suks on technical discussion

get on to Orkut i may answer it there

not worth reading

thanks for two points

Answer 5

Wow...

couldn't get that. Pls repeat your question ;)