Archive for the 'Opinion polls' Category

Under: Opinion polls
19 Oct 2016

 

Now I know that this is not the most interesting subject but I had of dose of this in college statistics classes (3) for my undergrad and graduate degree. In business, statistics are mostly used for continuous process improvement or validating supply and product variances. But I would dare to say that the way most of us know about polling is during the political seasons. :(

 

In this article I will give an overview of opinion polls and the parts that might make a difference in determining accuracy. While I am no expert, I have been exposed and understand the concepts. I will have one really good link that will explain this very well and three videos that are nice visual aids.

 

One of my pet peeves with opinion polling is that the confidence interval is not published as the sample size and margin of error. Typically one can have anywhere from 90% to 99% confidence interval – a huge difference this can be as important or more important than the margin of error! And this information is not easily available. But the media perpetuates the myth that the margin of error is the only (or certainly implies it is the main) factor to look for in polling data.

 

Here are the parts of a sample:

  • Survey questions
    • not too complex
    • not bias
  • Methods to reach sample population
    • landline phone
    • mobile phone
    • in person
    • via internet
  • Random survey of sample population
    • you cannot select the people – that is not random (for instance all subscribers of a liberal newspaper)
    • does it include ethnic, religious, young, middle aged, old, and other variances in the population?
    • If it is not random or not representative of the population – it is not valid (sample error)
  • Margin of error
    • How likely your samples will answer how you predict in the range
    • a +/- 3% margin of error has a 6 point range that likely can overlap with some of the other candidates
  • Confidence interval
    • How confident your survey results range will be be as predicted
    • a 90% confidence interval says that 10% of the time – the results can fall outside of the margin of error

 

Here is an good article entitled, “Why polls can be wrong…” and below are the videos if you like your information delivered audio/visually.

 

 

Full Story »