The following video illustrates how the randomisation test can be simulated with iNZight software.
INZight illustrates the randomisation test in its VIT module and the simulation is more dynamic
compared to the NZGrapher illustration.
Do not confuse the radomisation test with the random selection in sampling process.
Critical Vocabulary
  • Independent Variable- the part of the experiment that is controlled or changed by the experimenter
  • Dependent Variable- the part of the experiment that is observed or measured to gather data; changes because of the independent variable
  • Control- standard of comparison in the experiment; level of the independent variable that is left in the natural state, unchanged
  • Constant- part of the experiment that is kept the same to avoid affecting the independent variable
  • Hypothesis- educated guess or prediction about the experimental results
  • Qualitative observation- word observations such as color or texture
  • Quantitative observation- number observations or measurements
  • Inference- attempt to explain the observations
The difference of median months for baby's walk is the ESTIMATE (Fathom calls it a Measure) which changes (variability due to sampling). In the stop motion video there are two arrows or sticks. The longer one shows the difference of medians between the control and treatment groups. The shorter stick shows the difference of medians after one of the randomisations. Just imagine if we had 1000 of these differences (converted to numerical figures) and plotted them into a distribution, that will give us the final answer: 'is chance alone playing a role?'. That's how I see it, is it only a 'gamble' or is there a real difference?. Then only we will be able to say if the experimentation is effective or not.
What is placebo?
Did you find the answer? Or did you give up?
Above guidelines are given only as a reference. This is to clarify why some of the sources state the cut-off point at 10% . There are sources who would go by 5%. These are just guidelines. Students are advised not to follow them strictly. For level 3 assessment we will use roughly 10% or 0.1, as the 'no evidence' justification is more convincing that the null hypothesis does not hold.
Reference:  Cast, Massey University site ( the interpretation of the p-value).
Next provided link will get us a better picture of 'how 'significant' is this significance:
Note: This is a compilation work for my teaching purpose. Some are created by me, otherwise I followed the rule of  the Creative Commons.