Register now or log in to join your professional community.
In fact same problem I'm having right now. The only thing which we can do, perform descriptive analysis and bifurcate data in lessor or accurate data & data with errors. In your case to draw conclusion you have to work with 60% of your dataset which has high accuracy. Rest of the data, which is 40% of your original dataset, you have to apply different filters to test if any correction is possible in data. If possible make corrections, and compare the result of your analysis with accurate data set's results.