Experimental design principles

Students at this level are required to conduct experiments and need to be able to understand and use experimental design principles. They need to carefully work on the problem posing and planning stages of the statistical enquiry cycle to ensure that the experiment they design will generate the data they need to answer the problem posed. The aim of experiments is to show cause and effect.

Students need to define experimental units – these will be the people/objects that will be used for the experiment.

Students need to identify and describe the explanatory and response variables.

Students need to identify the key variables to be used as treatments - these should be clearly defined in the problem, but if a control is to be used this needs to be identified.

Students need to decide on methods of measurement that will minimise measurement bias.

Students need to identify other variables that might cause variation in the response variable.

These variables can either be controlled or not controlled for the experiment.

The plan should explain how the variables that can be controlled have been controlled in the experiment.
The variables that can not be controlled should be balanced by randomly assigning experimental units to treatments.

Students could consider blocking on a variable for the experiment if they have reason to believe that a particular factor may have considerable effect on the response variable. This involves splitting the experimental units into blocks, and then randomly assigning the treatment within each block.

Experimental design exemplar

Investigative problem

“Do Year 13 students at this school tend to remember more nouns than non-nouns?”

Plan

Four Year 13 Statistics classes will be used as the experimental units.
This is a total of 101 students.

The explanatory variable will be type of word.
There are two treatments for the explanatory variable: noun and non-noun

The response variable will be number of words correctly remembered.

The number of words recalled will be scored by an independent marker to minimise measurement bias.

Students will be randomly assigned either the list of 15 nouns to memorise, or the list of 15 non-nouns to memorise. This will give us the two treatment groups.

By randomly assigning the lists to students, other factors, such as individual student’s mastery of memory techniques, which may cause additional variation in the response variable, should be balanced across the two treatment groups. All other aspects of the experiment should be the controlled e.g. all students complete experiment at the same time of the day, all students have no noise present in classroom when completing experiment, all students have the same test conditions (1 minute to memorise the words, 2 minutes to recall the words) etc.

If there was evidence that girls had better recall for words than boys, we could investigate this factor by blocking on gender. The experimental units would be split two blocks: boys and girls. Within the boy block, each boy is randomly assigned either the list of 15 nouns to memorise, or the list of 15 non-nouns to memorise. Within the girl block, each girl is randomly assigned either the list of 15 nouns to memorise, or the list of 15 non-nouns to memorise.

Last updated October 5, 2011

TOP