Stories. Questions. Different perspectives.
Browse WAURE
← Advice & Questions
Career / Work · Advice & Questions ·

My manager wants an awkward result left off the chart. What would justify excluding it?

Career / Work category photographCareer / Work

One result in our materials test sits far from the others. My manager has asked me to leave it out of the summary chart because it is “obviously a glitch” and distracts from the comparison.

I checked the recorded value against the original file and it matches. I haven't found a documented instrument problem or a mistake in how that run was carried out. I also don't know enough to say the result is definitely meaningful. The raw records are still intact.

How would you ask for a sound decision about that point before issuing the report? I want to distinguish a justified exclusion from removing something because it makes the chart inconvenient, without accusing my manager of bad intent when there may be an explanation we haven't established yet.

Discussion

5 community replies

The NIST outlier guidance distinguishes verified erroneous data from unusual observations whose meaning is uncertain. An extreme point can warrant investigation without automatically warranting deletion. Ask what evidence or agreed analysis rule supports excluding this run; don't select a statistical test solely because it produces the preferred chart.

Check the run conditions with the person responsible for the method: instrument records, preparation and any recorded departure from the procedure. You've checked transcription, which is useful, but it doesn't settle every possible cause. Record what is known and what is still uncertain.

Ask how the report should describe an unresolved result and who approves that explanation. A transparent note or further investigation may be more useful than an apparently tidy summary. Preserve the original records and document any analysis change and its reason.

If a genuine equipment fault is found, would you still insist on displaying the point alongside the valid runs? That could confuse the reader too.

No blanket rule requires treating a demonstrated error as a valid observation. Correct it or exclude it through the appropriate documented method, and explain the treatment where relevant. The issue here is the missing justification, not a commitment to keep every point regardless of what the investigation establishes.

Join the discussion

More in Career / Work