Protocol (natural Sciences) - Calculations, Statistics and Bias

Calculations, Statistics and Bias

Protocols for methods that produce numerical results generally include detailed formulae for calculation of results. Formula may also be included for preparation of reagents and other solutions required for the work. Methods of statistical analysis may be included to guide interpretation of the data.

Many protocols include provisions for avoiding bias in the interpretation of results. Approximation error is common to all measurements. These errors can be absolute errors from limitations of the equipment or propagation errors from approximate numbers used in calculations. Sample bias is the most common and sometimes the hardest bias to quantify. Statisticians often go to great lengths to ensure that the sample used is representative. For instance political polls are best when restricted to likely voters and this is one of the reasons why web polls cannot be considered scientific. The sample size is another important concept and can lead to biased data simply due to an unlikely event. A sample size of 10, i.e. polling 10 people, will seldom give valid polling results. Standard deviation and variance are concepts used to quantify the likely relevance of a given sample size. The mass media and the public often use average and mean values interchangeably, which can lead to dubious and even misleading arguments. The placebo effect and observer bias often require an experiment to use a double blind protocol and a control group.