The question of analytics has popped up more and more these days, both from an organic and a paid perspective.
There are several analytics packages on the market today ranging from the very cheap to the not so very cheap; and from the very good to the not so very good as these things are wont to be.
Transitioning from one analytics package to another can be a jarring experience as numbers may fluctuate wildly between any two given packages.
There are any number of reasons for that. One package may consider visits as time spent with increments that vary (15 mins versus 30 minutes). One package may try and differentiate between bot or automated programs and human use. Some may track through flash sites better than others. Caching may be an issue.
A side by side comparison will generally show variance in results.
Ideally, for a while, ranging anywhere up to 3 months, both packages should be run in concert if you're transitioning your analytics package. It will help to create a baseline from which you can get a better sense of how your new analytics suite works.
Moreover, if you have issues with your new analytics package and make the choice to disable or reinstall, you haven't compromised or, worse, lost valuable site history.
With any analytics package, try and ensure a full year's worth of data before switching products. This will allow you to have a better sense of seasonality and annual trends. (By example, summer traffic tends to differ dramatically from winter traffic which, for ecommerce sites tends to differ dramatically from the Christmas shopping season.)
For a great (two part) article on things that can through your stats, Brandt Dainow covers the subject well here.
Have a good one.
~ The (SEP) Guy