Could you round to buckets as well - take the 10 (say) most common screen sizes, and round users to the nearest? That way users with odd screen sizes aren't ide

Could you round to buckets as well - take the 10 (say) most common screen sizes, and round users to the nearest? That way users with odd screen sizes aren't identified.