There are lots of interesting ways to make statistics come out the way you wish, just ask an old social scientist like me. This American Thinker article reports on how the Associated Press has biased the numbers to make President Obama look good at the end of his first 100 days. The article links you to AP articles which claim that 48% of Americans think the country is on the right track.
Steve McCann, who wrote the Thinker article, determined that the poll surveyed almost twice as many Democrats as Republicans. Actually, Obama won 53% of the popular vote while McCain won 47%. So...how can the AP justify polling nearly twice as many Democrats as Republicans? Do they think lots of Democrats voted for McCain? It appears AP believes all 18% of those who claimed no party identification were voters too embarrassed to admit they were Republicans.
Just suppose they'd surveyed almost twice as many Republicans as Democrats. What would the "right track" percentage be? You'd guess it would be much lower, and you'd be right.
The right way to do this poll is to select for your sample percentages of Republicans and Democrats on a rough parity with how many voted each way in the most recent election. Another defensible way is to make your respondent percentages equal each party's registrations. Very clearly, AP did neither, probably because they didn't like the results.
