172 views
edited Jun 17, 2017

I want to know whether we need to apply continuity correction when we use normal approximation to carry out a statistical test. Continuity correction is applied in ActEd but not mentioned at all in lecture slides.

For example, in a sign test, if the no. of positive $$Z_x$$ is P which is greater than $$0.5*m$$ (no. of age groups). Thus we want to test whether P is too large and we use normal approximation to do so.

The test statistic should be $$\frac{P-0.5m}{0.5\sqrt{m}}$$ or $$\frac{P-0.5-0.5m}{0.5\sqrt{m}}$$ (i.e. continuity correction)

Just want to be sure since it is not mentioned in the lecture slides.

If you use a normal approximation, then you should use a continuity correction, if the number of age groups is not large enough ($$\lt 20$$).