Let's say we have a marketing company taking a survey on a population N on each of the 50 states. The marketing company tells the statistician that they will send two samples, one with 70% of the polls and then the rest of the polls.
The poll is simple, which product does the population prefer: Product A, Product B, Product C. The total sampling is 10,000 surveys for each state (N=10000). For simplicity's sake we will only take into consideration three states:
70% of the polls computed (so 7,000 for each state)
Florida
Product A: 42%
Product B: 38%
Product C: 20%
New York
Product A: 53%
Product B: 40%
Product C: 7%
California
Product A: 55%
Product B: 40%
Product C: 5%
Now, the marketing company asks a simple question. If the the remaining 30% (the rest is 3,000 surveys for each state) is computed, how is each product's percentage expected to change foe each state? (what range +/-, stdv maybe?, so how do we calculate the standard deviation?)
So summarizing... The marketing company wants to know one thing: "Is the outcome expected to change or stay the same after the rest of the surveys are computed?" Can we assume that 7,000 is a good estimate of the whole population considering the TOTAL population (of people that tried the three products) is 10,000?