I have a range of data where some of the values are reported as < or > ( e.g. <0.5%, > 99.9% etc). Due to < and > sign, I cannot change the column type to 'numeric'. Hence I cannot do any data analysis like std. deviation or plotting chart. etc. Is there a way to eliminate those signs and keeping only absolute values?
@txnelson showed you how to do what you asked, but is that what you really want to do? Cases where observations are beyond the limit of detection (LOD) or limit of quantitation (LOQ) are common but often analyzed incorrectly.
The instances that you want to change are censored observations. They are not the actual value, but a lower or upper bound on the actual value. Using the bound (e.g., .05%) for the value will bias your statistical analysis. Using the limit (e.g., 0%) for the value will bias your statistical analysis. Using the mid-point (e.g., 0.025%) for the value will bias your statistical analysis. There are procedures in JMP and JMP Pro that properly analyze censored data with maximum likelihood estimators (MLE) to produce unbiased results.
You would need to use interval censoring in the case of a lower bound. Here is an example of how it might be set up:
The procedures that address censoring recognize the Lower and Upper data columns together in the Y role as interval censored data.