Re: Normal distribution of ratio of means

Thanks for your helping remarks.

I remain with one question to each of your two last sentences:

1)"If the probability that the denominator mean equals zero is very
small, then using the normal approximation for the ratio does lead to
pretty accurate confidence intervals."

Could you please give me a reference for this statement (since I have
not been able to find one myself; I had also a look at Hinckley's "On
the Ratio of Two Correlated Normal Random Variables", Biometrika, 56,
635-639, but could not find there such a reference; it might be that
that there is such a reference in this - quite complicated - article,
but that I do not see it)?

2)"Be sure to account for the correlation between numerator and
denominator when figuring the standard error of the ratio."

Does this mean that there must be a correlation between numerator and
denominator (and if yes: Is it given in the case I described?), or does
it simply mean that I should state the coefficient of correlation
between nominator and denominator?

.

Quantcast