Talk:Observational error

[Untitled]
What is meant by Timing Error terminology in Management —Preceding unsigned comment added by 210.56.13.83 (talk • contribs)

Merger proposal
I think this article and Approximation error are talking about the same thing. Should we merge them? --Surturz (talk) 02:52, 4 March 2009 (UTC)


 * They aren't the same thing. Observational error relates to taking a measurement.  Approximation error includes other types of errors - for instance those introduced by using approximate values (e.g., any value used for PI in a digital numerical calculation will be an approximation), those introduced by ignoring less significant effects in a computation, etc.  So some observational errors may be examples of approximation errors, but things like systematic measurement error are not particularly relevant to approximation error.  Think they should stay separate. Zodon (talk) 06:12, 4 March 2009 (UTC)


 * I agree, they are clearly distinct topics. Approximation error needn't have any random component, but randomness is pretty fundamental to the concept of measurement/observational error. The articles should be expanded to make the distinction clear though. -- Avenue (talk) 07:07, 4 March 2009 (UTC)
 * I will remove the merge tags then--Thorseth (talk) 08:30, 14 May 2009 (UTC)

Non sampling error
I thought it was rather odd that Non_sampling_error redirects to this page (Observational_error), and not Non-sampling_error...... — Preceding unsigned comment added by 130.216.51.121 (talk) 00:33, 24 September 2012 (UTC)
 * Now fixed as suggested. Melcombe (talk) 23:24, 12 April 2013 (UTC)

New merge proposal
I oppose the proposed merge with Systematic error and Random error. Those two topics deserve their own articles. -- 202.124.73.40 (talk) 08:43, 3 June 2013 (UTC)


 * Both articles contain sections systematic versus random error therefore there is substantial duplicated material. Fgnievinski (talk) 01:16, 29 June 2014 (UTC)


 * ✅ They're so strongly related it's easiest to discuss them by contrasting them. They're both quite short articles, so there's no danger of excessive clutter. 71.41.210.146 (talk) 10:03, 26 September 2014 (UTC)

I do not agree with a sentence
I do not agree with the sentence "The higher the precision of a measurement instrument, the smaller the variability (standard deviation) of the fluctuations in its readings.". Take for example a 1m long object and a 1m ruler that has no divisions, i.e. its precision is 1m. One can say it has a very low precision. Yet someone using it will always measure the object as a 1m object. There is no fluctuation in its reading. Now if you make the ruler more precise by creating, say 1000 divisions, then you will introduce a variability of the fluctuations in its readings. In this case, with a higher precision of measurement instrument, the higher the variability (standard deviation) of the fluctuations in its readings. This contradicts the current article's sentence. (unsigned comment). I think the matter can be clarified by differciating between statistical (randomness) error and reading error. If you have a ruler divided by a millimeter scale, the minimal difference readable is by 0,3 mm. When you measure the length of an object by a mm screw you can distinct 0,1 mm distances. Measuring multiply the object you obtain readings, that may differ by up to 0,4 mm. From this you obtain a mean and a statistical uncertainty, that diminishes the more readings you make. But on the other hand, each reading has a reading uncertainty by 0,1 mm. And this reading error applies also to the mean. So, when the statistical error goes under 0,1 mm, further measurements are useless, as the total error never will never be less than 0,1 mm. Additional note : Less than 10 readings are useless for a statistical evaluation. (Dok21fie (talk) 06:14, 25 March 2019 (UTC))

No mention of Bias Error
I'm disappointed that there's no mention of bias error in this wiki.

Bias error is the error purposefully induced by an observer and motivated by a desire to confirm or deny a hypotheses and/or to avoid more work (i.e.: accepting a slightly negative result as a positive one, so the experiment need not be repeated.)

While the referenced wiki for Systemic_bias touches on it, that wiki mentions it as a consistent occurrence, while bias error most often occurs at the level of an individual observer, and occurs sporadically. TCav (talk) 16:25, 15 January 2018 (UTC)

Bad example
The thermometer measuring -100 as -102, 0 as 0, and 200 as 204 is a confusing example of percentage error, because 0 is just an arbitrary point on the temperature scale, and there is no reason for the thermometer to be exactly accurate at that point. A better example would be a measuring tape that gives a measurement of 10.2 meters instead of the correct 10, 20.4 instead of the correct 20, 5.1 instead of the correct 5, 15.3 instead of the correct 15 and gives 0 when measuring a distance of 0 (in that case the points are in the same place, and the tape is unnecessary). I will edit the page to address this if there is no objection. 882,614,759edits (talk) 16:48, 2 March 2018 (UTC)

Systematic error
I don't believe this is correct:
 * If the cause of the systematic error can be identified, then it usually can be eliminated.

Surely eliminated implies a perfectly calibrated instrument, and the most we can do is make the systematic error negligible compared to the random error for a set of readings or to the precision with which the instrument can be read. And even a systematic error smaller than the divisions of the instrument will affect the point at which the reading changes from one division to the next. Musiconeologist (talk) 19:36, 30 March 2024 (UTC)