# Determining '< x' values as numeric

I do have a large dataset containing nearly a decade of particulate matter measurements. At the lower precision limit of the measuring instrument, there are values given like < 8. For statistical analysis however, a defined value is needed.

What is the best possible way to deal with this?

• Replacing it as missing values?
• Replacing it by 8 or even 7 (< 8)?
• Assuming 0?
• Assuming the mean value between something like 0 and 8; e. g. 4?

Or is there a better solution, maybe considering the natural background value?

EDIT: I'd like to calculate the intra-day- and intra-year distribution, as well as searching for patterns by considering wind directiond and speed.

• It depends on what you want to measure/calculate. So we need more info. – Jan Doggen Oct 3 '16 at 10:05
• In addition to Gordon Stanger's comment: replacing <8 values by missing values might alter the distribution but replacing < 8 values by noise (or replacing it by other non-missing values) might yield artificial correlations. Therefore, it might be reasonable to do the noise and the missing-value approach and get a feeling on which one is actually working 'better'. If there are less than approximately 5% (not a fix threshold) I would suggest to use missing values. – daniel.neumann Oct 4 '16 at 7:45