<p>A quick stat-related question...say you have a preference poll between orange juice and fruit punch. Say it's .8 oj, .1 fruit punch, and .1 undecided proportions in a sample. If you're interested in the difference between the proportions in oj and fruit punch, could you do a one proportion z-interval with .7? Why or why not? Also, does the undecided play any role in this? Finally, what would you need to do to use a 2 prop z-interval for the same difference?</p>
<p>This is a difference in proportions, so it would have to be two-sample interval and you construct your confidence interval around the difference, unless I’m going crazy here. Also if you are just doing the confidence interval I don’t think you need to specify one or two-sample – you only do that for significance tests, right?
The undecided does not play a role unless it’s mentioned.
The one-prop .7 and two-prop interval are the same thing…</p>
<p>Whoa my post got really messed up. Hmm but yeah, there is no difference and you would build it around .7 either way.</p>
<p>you would use .7
because its your p represents what you want, which is .8 for oj, and your q represents what you don’t want. you’d combine the two (fruit punch and undecided)
so it would be .8-.1</p>
<p>What are the hypotheses?</p>
<p>About the 2 prop z interval…I don’t think the two proportions are exactly independent here…Should I take another sample and use the fruit punch proportion of that? Is it pointless to even do this kind of difference? I am confused,</p>