Increased statistical power

Isogenic strains  tend to be more uniform than outbred stocks. This is due  to the absence of any within-strain genetic variation. This is important because it means that fewer animals are often neededin an  experiment when using isogneic strains  than when using  outbred stocks.

The table shows the mean and standard  deviation of sleeping time under hexobarbital anaesthetic in five inbred strains and two outbred stocks of mice. Note that the standard deviation is much larger in the outbred stocks.

Suppose an experiment is to be done to study the effect of some treatment on sleeping time. The mean sleeping time in a treated and control group are to  be compared using a two-sample t-test. Using statistical power calculations (see companion web site) it is possible to estimate the number of animals which will be  needed, assuming the aim is to detect a change of 4 minutes of more using a 5% significance level and a 90% power. This is shown in column 5. Using inbred mice, sample sizes of 7-23  mice per group (depending on strain) would be needed, whereas if outbred mice are used groups of 191 to 297 would be needed.

Alternatively, if the sample size  is fixed at 20 mice per group, then the power of the experiment to detect a four minute difference in mean sleeping time between treated and control mice  is shown in column 6. It would range from 86  to 99% using inbred mice or 13 to 17% using outbred mice.

These calculations assume that the response to any treatment is similar in inbred and outbred stocks. While there is no evidence that the two classes are likely to  respond differently, individual strains and stocks may do so.

The table explains how control of variation leads to more powerful experiments

Table 1. The importance of phenotypic  uniformity in determining sample size. Hexobarbital sleeping time in mice (see text for explanation).





Sample size(4)
































CFW (outbred)






Swiss (outbred)






(1) Number in each group

(2) Mean sleeping time

(3)  Standard deviation of sleeping time

(4) Number needed in a two-sample t-test to detect a 4 min. change in the mean  (2-sided) with a 5% significance level and a power of 90% 

(5) Power  of an experiment to detect a 4 min. change in the mean if the sample  size is fixed at 20 mice/group 

Data from  Jay 1955 Proc Soc. Exp Biol Med 90:378