Originally Posted by
galactus
If the null hypothesis is not true, then the alternative hypothesis must be true.
Important to note that this only holds if the hypotheses are mutually exclusive AND cover all the possible subsets. It is a common mistake made by novices to stats and other scientists.
More importantly, in stats, rejecting the null is almost never an all-or-none proposition. That is, as you have put it "If the null hypothesis is not true", is usually actually phrased as "To x degree of certainty", or something like that.
To further explain this, if my p value is less than x (usually 0.05 or 0.01) then I reject the null. That doesn't mean that the null hypothesis is not true. Rather, it means that 95% (or 99%) of the time that this threshold is exceeded, the null hypothesis can correctly be rejected. The rest of the time (eg. 5% or 1%) even when the null hypothesis is true this threshold will be exceeded, and so it may be falsely rejected.
Well, I could go on about this for hours, but I think we have gone way farther than the questioner really needed.