Simple probability. If 2/3's of any group has trait X the chance you will run into X increases the smaller the group is.
Err, no.
The likelihood of finding trait X in a total population of Y will be X/Y
regardless of the initial size of that population. Of course the actual likelihood will change if the population changes, although the calculation for that likelihood remains constant.
Likelihood of X = number of X in population/Total Population
How it 'seems' is
irrelevant.
In your case of M&Ms it would make no difference if you started with 100 or 100 million. The
relative numbers of each colour would remain the same. Thus the count of each colour removed would be the same (well up to the number in the smaller sample)
assuming they were removed at random from their respective populations according to the initial statistical probability - 2 green for each red, the ratio would
always remain 2/3. Of course that's unlikely, so the probability of a green would vary within and between populations - but the
actual size of the populations has
no bearing on this - and over time it would even out.
Your other examples are borderline incomprehensible, and flawed. For example:
Since people are social animals they tend to hang around those with the same traits, another variable, so if there were 1,000 people on the website and let's say 900 were not 100% straight and 100 were dividing both in half that means 450 people out of the non-selected group will be desireable and only 50 out of the 100% group will be desireable.
No, it doesn't. Well, it might, depending. Your 'random' split theme suggests that 50% are desirable, in which case - the dividing in two bit ... it's meaningless, the likelihood of someone being desirable is 50%
regardless of the 'split'. The numbers in each group are irrelevant.
If the criteria for selection i.e. 'desirable'
was being 100% straight, then again your example makes
no sense. If you isolated the 10% that
are 100% straight as desirable
based on that criteria you will now have
two separate populations, one of 100 and one of 900. The desirables in one group will be 100/100 or 100%, and 0/900 or 0% respectively.
Alternatively, you divide the initial population into two groups at random then
each new population will each contain 10% desirable and 90% undesirable. Dividing each new population in half again, won't change the likelihood of picking a 'desirable', it will remain 10%.
If your underlying criteria for desirable is simply 50% then again, these splits make no difference, it will remain 50%. You seem to be confusing sample size with probability. Alternatively you're not explaining your thinking clearly enough for me.:biggrin1:
Also it
could be me, I
hated probability and it's late.:frown1:
Sorry about diversion. I can't say I've noticed a trend, the idiotic seem quite evenly distributed.