The data isn't hugely rich, so there's only a certain amount of insight you can get. If you had more data (eg, number of attempts at each level, time spent on each level).
In your position, the main question I would be trying to answer is: Which levels cause an unusual number of players to stop playing? I suspect that simply plotting the data you have onto a chart would help you spot the answer without any further calculations.
I would do this by calculating the users lost for each level:
UsersLost(level) = Users(level - 1) - Users(level)
(Depending on the shape of the data, you may find it more useful to model the proportion of users lost, rather than the absolute number.)
I would look at the distribution of these numbers to help me determine the next information I was interested in. Again, plotting the results is a useful exploratory tool.
Ultimately, I'd be looking for any levels where there's a big difference between the expected number of users lost and the actual number of users lost. These are the levels you want to investigate closely, to help identify the qualitative reason that they lose/keep users.
At this stage, I think that you'd be better off asking on a forum more suited to exploratory analytics and the user experience -- someone suggested gamedev.stackexchange as a good starting point. Mathematics might be more useful when you have figured out what you're looking for (and need help finding it).