Intrinsic problems in the digitizing process

[danieleongari]

I report here common problems that may emerge in the digitizing, which add noise to the data:

LEt us consider Fig 4A from 10.1126/science.aaz8881:

The results of my digitizing of the curve @ 296K is:

One can see two problems, when my data are zoomed:

• the low pressure point is negative because it is hard to pick it correctly. It would be better to set it to 0,0 or remove it
• since the curve is skewed, a small error in picking the marker results in a large error compared to the max uptake and a curve that looks odd. Indeed, since it is common to report isotherms at different temperatures in the same plot, we have to take for granted that this problems will generally raise for high temperature isotherms.

[dwsideriusNIST]

My standard cleanup scripts do the following:

1. Remove all points with p < 0
2. Set Ads = 0 for p>0 and Ads<0

The second problem is an inherent issue with digitizing graphical data and you are correct that this is amplified in plots with multiple temperatures. The other time this is common is when gases at different conditions relative to their critical temperature are plotted in a single figure with common y-axis (e.g., single material, but CO2 and N2 at 300K; the N2 uptake will be very low compared to CO2).

[ltalirz]

The two rules @dwsideriusNIST mentions could easily be applied at the "plot" stage - just wondering whether this should be done by the digitizer or at the point of ingestion to the DB.

@dwsideriusNIST any thoughts?

[dwsideriusNIST]

The two rules @dwsideriusNIST mentions could easily be applied at the "plot" stage - just wondering whether this should be done by the digitizer or at the point of ingestion to the DB.

@dwsideriusNIST any thoughts?

I would accept a widget in the plotting tab of the app that applies those rules, but it needs to be initiated by an active decision of the user, not passively in the background - and warn the user beforehand. Essentially, make a big fuss to communicate to the user to "take a good look at your data before applying these rules."

My post-processing, pre-ingestion scripts will still apply these rules