Page:Sm all cc.pdf/117

From Wikisource
Jump to navigation Jump to search
This page has been proofread, but needs to be validated.
114

[Image removed for copyright reasons] (Wikisource contributor note)

12) use identification labels on samples. Use indelible ink for labeling, and assure that either labels cannot be lost or labeling is redundant. If samples are in containers and if feasible, label both the sample and container.

13) take methodical, detailed notes during the experiment.

  • Do not trust your memory for anything. Remember that you may come back to these notes months later, long after short term memory of temporarily obvious factors has faded.
  • Do not use scraps of paper. Ideally,use a lab notebook; some researchers say that a lab notebook is essential. At least, use dated sheets of paper and either a 3-ring binder or manila folder.
  • Sketches may be useful.
  • Decide whether or not to use a checklist during the experiment.
  • Prepare and use a standard form for routine measurements, both to facilitate later analysis and to assure that all relevant information is recorded.
  • Note times of steps, sample ID’s, experimenter (if more than one), and anything else that remotely could be considered a variable (e.g., source and grade of chemicals) in later review of the experiment.
  • Note units of all data. A frequent pitfall is to assume that the units are so obvious or familiar that you could not forget them.
  • Note any changes to the experimental procedure or equipment.
  • Document calibrations.
  • Record raw data, not just corrected data, because you may decide later to use different corrections. Record correction equations, because you may wonder later whether or not you did all corrections properly. Raw data are better than corrected data, if the corrections are untrustworthy or of unknown accuracy. For example, in using ‘temperature compensated’ equipment, I have been confronted with the challenge of evaluating whether the compensation actually introduced error because of its inaccurate measurement of temperature.
  • Record bad data, unreliable results, and abortive experiments, using obvious flags to avoid mistaking them for trustworthy data (e.g., draw a large X through them). Add a notation on why they failed. Possibly, later analysis will show that information can be salvaged from these discards. Certainly, one wants to minimize the chances of making the same mistake twice.