Randomize the order of runs

The consequences of an experimental error on the conclusions drawn from experimental data were discussed in Chapter 3. To allow this error to be analyzed by known statistical probability distributions, such as the normal distribution, the t-distribution, the F-distribution, certain assumptions as to the experimental error must be made. These assumptions are  

However, chemical systems can be influenced by disturbances which are not random and which will produce a systematic error, e.g.  

There is always a risk that the experimental result may be influenced by non-random, time-dependent errors. Such risks may be counteracted by precautions taken by the experimenter, and the remedy is randomization. This means, that in any situation where the experimenter has a choice as to in which order he/she should do things, the choice should be a random choice. For example, the order of executing the experimental runs should be randomized the order of analyzing samples drawn from the reaction mixture should also be randomized if several samples are analyzed on the same occasion. 

These precautions will break time-dependent phenomena, and transform systematic errors due to such events into random errors. 

A random order of experiments can be established in several ways, e.g. by writing an identification number of the experiments on separate pieces of paper, putting them in a hat, and then drawing them at random, by using tabulated random numbers or by using the random number generator of a computer. Personally, I use a well-shuffled pack of cards. 

---