Difference between systematic and random errors

1
3. WHAT IS THE DIFFERENCE BETWEEN SYSTEMATIC & RANDOM ERROR?a) SYSTEMATIC ERRORS/UNCERTANTIES(affect accuracy) Affect ALLmeasurements in the SAMEway [i.e. measurements that are either consistently too high or too low...due to systematic error(s)]REPEATEDmeasurements WILL NOTreveal this type of uncertainty regardless of the number of trials performed.Usually difficult to detectunless the expected/actual value is known.Can sometimes be remedied/ameliorated/lessened.Due to: a)Miscalibrations / “faulty” or damaged equipment Examples:i)A stopwatch running 2 seconds too slow.*Repeated measurements will not reveal this uncertainty.*Remedy= check clock against a more reliable one (If doubt the reliability of a measuring device try to check it against a device that is more reliable.)ii)A miscalibrated pH metre will always display a reading that is either higher or lower than the actual value.iii)A thermometer reporting a temperature that is 2 degrees higher than the actual value.iv)A stretched ruler.ORb)Poor experimental technique (“human error”)Examples:i) Reading a meniscus incorrectly / parallax error? ii) Reading scale incorrectly iii) Unaccounted heat loss in a calorimetry experiment *A note on “human error”: Please don'trefer to "human error." Examples of so-called human error include misreading a ruler, adding the wrong reagent to a reaction mixture, mis-timing the reaction, miscalculations, or any kind of mistake. Scientists would never report the results of an experiment affected by human error -instead, they repeat the experiment more carefully!Points will be deducted from your lab report if you discuss "human error" instead of "experimental error."

b) RANDOM ERRORS/UNCERTANTIES (affect precision) Actual value may be higher OR lower than the value you record....in this manner, they are “random”...just as likely to overestimate as to underestimate a measurement.REPEATEDmeasurements WILL reveal this type of uncertainty.They are PREDICTABLE and UNAVOIDABLE.Random errors can generally not be ameliorated.Arise mostly from the INADEQUACIESor LIMITATIONSinherent to all measuring instruments. The degree of random error can be QUANTIFIED.The random error is equivalent to the uncertainty in measurement.This is usually given by the manufacturer of the equipment and expressed as +/ -a certain value. *If this information is not available, a good guideline is to estimate the uncertainty atHALF of the SMALLESTdivision on a scalefor an ANALOGUE device and the SMALLEST division for a DIGITAL device.Note when the uncertainty is recorded, it should be to the same number of decimal placesas the measured value. For example a balance reading to 53.457g +/-0.001(↑ decimal places = ↑ sig. digs. = ↑ precision)Example 1:Imagine trying to determine the period of a pendulum.Best strategy = Make repeated measurements; calculate average (...the best average of the period is the average value)*the mean is the best estimateof a measurement based on a set of measured values*standard deviation= ?e.g. Four measurements(in seconds) = 2.3, 2.4, 2.5, 2.4; average = 2.4Range = 2.3 to 2.5Example 2: multiple students take temperature of room (discuss both system and random error)Sources of random errors cannot always be identified. Possible sources:a) observational e.g. reading burette, judging a colour changeb) environmental e.g. convection currents*standard deviation and precisionWhen the final uncertainty arising from random errors is calculated, this can then be compared with the experimental error as described above. If the experimental error is larger than the total uncertainty, then random error alone cannot explain the discrepancy and systematic errors must be involved. Almost all measurements are subject to both systematic and random uncertainties.SUMMARY TABLE: SYSTEMATIC VS. RANDOM ERRORSSYSTEMATICRANDOM-affect results same way (either all too high or all toolow)-affect results randomly (some too high some too low)-not revealed by repeated measurements-revealed by repeated measurements-can sometimes be eliminated-can never be eliminated-often difficult to quantify (unless true value is known)-can always be quantified-cannot be treated statistically-can be treated statistically