# What we substract 1 twice

HI,

I understand that alpha is the confidence level, I don't see why we have to subtract 1 from alpha/2?

2 answers ( 0 marked as helpful)

for a one sided test (lets say we are testing positive side only) we want all of of the normal distribution to the left 50% and then all of the area to right up to alpha. If alpha was 1% the value of this marker would contain 99% of values to the left and 1% value to the right. In a two sided test we need to ensure we have 99% of the data in the middle of the distribuiton, 0.05% of data on the extreme left and 0.05% of data on the extreme right (hence alpha / 2).

the example here says that 2.5% values is less than 94833, 2.5% is greater that 105568 and 95% of values in between, hence the 95% confidence interval