Measurement error

[IvaBio]

I'm aware that this is a very simple question, nevertheless I'd like to get some help.

I'm evaluating the precision of measuring angles in a certain experiment - the object lies in a Petri dish under a certain angle to the 'north' position. I have measured the photo of the object 15 times during different 15 days, so I have 15 values (mean and median = 336.6 degrees) and range of measurement (335.5-337.0, i.e. 1.5 degree). Standard deviation of mean is 0.39 degree.

My question is, do I understand correctly that the error of measurement will be (SD/mean)*100% (i.e. 0.11%), and not (1/2*range/mean)*100% (i.e. 0.22%)?

Thank you.

 

[Miner]

It is not as simple as you think. Measurement error means different things in different disciplines. For example, in my discipline (industrial statistics), measurement error can be decomposed into:

• Accuracy
  • Bias
  • Linearity
  • Stability
• Precision
  • Repeatability
  • Reproducibility

With your data, you could potentially assess bias, stability and repeatability. In other disciplines, the answer will likely differ.

 

[IvaBio]

Thank you. May I ask, the calculated value (SD/mean)*100% will correspond to bias or other characteristics?

 

[Miner]

Your formula is the coefficient of variation converted to a percentage. It would be one approach for assessing repeatability of a measurement. However, it is not used in my discipline, so I cannot comment on its relative merits or flaws.

 

[IvaBio]

Thank you. I see that I have to consult literature on measurement errors in life sciences.
If anyone can provide references to valuable sources on this topic, I would very much appreciate this.

 

[fed2]

sounds like your data may be 'circular': that could turn out to be important. You can use box-plots.