The Gemini conundrum--a problem of unpairable data: statistical comparison of ratios each derived from two separately estimated parameters.

In biological studies it may be necessary to compare ratios of two separately estimated parameters under test and control conditions. Since it is not always possible technically to obtain paired data for the two parameters, how does one take account of their variances? By extending previously described approximations in a modified t-test we present a BASIC computer program which may be used to solve this problem.