In: Statistics and Probability
2. The managers of three different supermarkets in the US desire to test the effectiveness of a new store design on sales. The manager of supermarket A decides to let store directors choose on their own whether they judge appropriate to implement the new design. The manager of supermarket B randomly chooses stores to assign the new design. The manager of supermarket C randomly draws stores in each US State separately.
(a) State whether each of the three experimental designs is observational or experimental, providing an explanation. (7.5p)
(b) Which experimental design is implemented in each of the three studies? (7.5p)
(c) Which kind of bias will arise from design A? Why? (5p)
(d) Compare the quality of statistical results from Designs B and C.State which of the two prevails on your opinion and why. (5p)
A
The manager of supermarket A chooses experimental study as he lets store directors to choose whether they want the new store design. Here, the treatment group is all the stores chosen for the new store design and control group is the stores not chosen for the new store design.
The manager of supermarket B chooses experimental study as he randomly draws stores for the new store design. Here, the treatment group is all the stores chosen for the new store design and control group is the stores not chosen for the new store design.
The manager of supermarket C chooses experimental study as he randomly draws stores for the new store design from each state separately. Here, the treatment group is all the stores chosen for the new store design and control group is the stores not chosen for the new store design.
B
The manager of supermarket A chooses quasi experimental design because he does not randomly select stores for treatment.
The manager of supermarket B chooses pre experimental design as he draws randomly from the population to assign treatment group but no control group is drawn or selected
The manager of supermarket C chooses pre experimental design as he draws randomly from the population to assign treatment group but no control group is drawn or selected
C
Selection bias will arise from design A as the store directors would be having their inherent bias to choose new store design
D
If the stores are distributed equally across states and the purchasing power across states is equal, then both are expected to give same statistical results. But that scenario is highly unlikely and in that case, we should account for the difference in stores and purchasing power of different states. And hence, assigning the stores to two groups by randomly drawing from each state separately gives better quality of statistical results with variance accounted for.