data management

categorical (qualitative) data

Data that can be sorted into distinct groups or categories.

ordinal data

Qualitative data that can be ranked. Examples: poor, fair, good, very good

nominal data

Qualitative data that cannot be ranked.Examples: blue eyes, Green eyes, Brown Eyes

numerical (qualitative) data

Data in the form of any number.

discrete data

A type of data that consists of distinct, separate, and countable values that can't be divided into smaller parts.

continuous data

Numerical data that can take on any value within a given range and is measured rather than counted.

simple random

Randomly choose a specific number of people Ex: systematic and stratified samples

Put all the names of a population into a hat and draw one or several names out. This way, everyone has an equal chance of being chosen.

systematic

Put the population in a list that is ordered and choose different people at regular intervals.

Order the patients of a doctor in a certain way (e.g., alphabetically) and choose one randomly. At regular intervals, select the rest of the data starting from the original point. (eg. every tenth name after the original).

stratified

• Divide the population into groups with equal proportions, as those groups in the population • Time and cost-efficient to conduct

Survey factory employees about new safety initiatives: There are 1000 employees in the factory, of which 633 are women and 367 are men. Randomly select 63 women and 37 men to take the survey.

cluster

Divide the population into groups, randomly choose a number of the groups, and sample each member of the chosen group.

Survey Little League Canada baseball players. Randomly select five districts in each province and give the survey to every player in those districts.

multistage

Divide the population into a hierarchy and choose a random sample at each level.

Conduct an employee wellness survey by randomly selecting 10 stores. Randomly select three departments in each store, and randomly select 10 employees in each of those departments.

convenience

Choose individuals from the population who are easy to access • Can yield unreliable results since it inadvertently omits large portions of the population • Often very inexpensive to conduct

To get the public's input on a new pet by-law, a local politician goes to a local park and asks people their opinions.

voluntary

• Allow participants to choose whether or not to participate

• Often, the only people who respond are either heavily in favour or heavily against what the survey is about

• ex: Conduct an online poll asking people whether banning junk food in schools will fight obesity.

response bias

When respondents change their answers to influence the results, to avoid embarrassment, or to give the answer they think the questioner wants.

A teacher asks students to raise their hand if they cheated on last week's test. Students will not want to admit to cheating on a test so it is unlikely that many will raise their hand.

sampling bias

When the sample does not closely represent the population.

A politician goes back to the farming community she grew up in to ask for opinions on her latest initiative for the agriculture industry. It is likely that a larger proportion of the people she speaks to would support the initiative, both because it would benefit them and because she grew up in the area. This would not accurately represent the entire population.

measurement bias

When the collection method is such that the characteristics are consistently over- or under-represented.

A survey asks, "A lot of people do not like math. How would you feel being referred to as a math geek?" This is a leading question; the wording of the question can affect the outcome by influencing someone's answer. Other types of measurement bias can occur when the collection method affects the results, for example, when the options in a multiple-choice question are too limited for an honest response.

non-response bias

When the opinions of respondents differ in meaningful ways from those of non-respondents.

A mail-in survey asked respondents about their drinking habits. Only 3% of the surveys were returned. Such a small return rate would likely not yield a representative sample. In fact, those who respond often have very strong opinions about the subject matter, and so the results could easily over- or under-estimate the feelings of the population.