POL242 – Tuesday October 23 2012
EKOS polls are conducted using Interactive Voice Response (IVR) technology, which allows
respondents to enter their preferences by punching the keypad on their phone, rather than
telling them to an operator.
In an effort to reduce the coverage bias of landline only RDD, we created a dual landline/cell
phone RDD sampling frame for this research. As a result, we are able to reach those with a
landline and cell phone, as well as cell phone only household and landline only households.
This dual frame yields a near perfect unweighted distribution on age group and gender,
something almost never seen with traditional landline RDD sample or interviewer administered
The field dates for this survey are October 3-11, 2011. In total a random sample of 2,391
Ontario residents aged 18 and over responded to the survey (including a sub-sample of 2165
decided and leaning voters). The margin of error associated with the total sample is +/- 2.0
percentage points, 19 times out of 20.
Please note that the margin of error increases when the results are sub-divided (error margins
for sub groups such as region, sex age, and education) All the data have been statistically
weighted to ensure samples composition reflects that of the actual population of Ontario
according to Census data.
Margin of Error
• AKA “Sampling Error” or “The standard error of proportion”
• Many ways of creating errors in sampling (dumb questions etc…)
• Standard equation for determining margin of error = 1.96 Square Root p(1-p) over
1.96 p 1−p )
• 1.96 = number of standard deviation on a normal curve (in 95% of cases). We are talking
about 95% of the time vs. 5% of the time. 5% = 1/20. So therefore, it is 19/20 times as a
reference to the standard deviation on a normal curve.
• P(1-pI). This demonstrates the most variation you can have. (.25->.50)
_____ N This demonstrates the number of cases given.
Types of error
• Type one error: when I think there is a relationship but there isn’t
• Type two error occurs when I think there is no relationship, but there is one.
• Statistical Significance implies you can rule out the null hypothesis. Statistical
significance has very little do with substantial significance.
• All levels of signi