Week 07 Flashcards
Numerical scales for measuring
Numerical value must identify scale on which measurement it made (days, metres, kg, degrees celcius)
Non-numerical scales
Within a category:
- countable (quantity)
- in categories (male female)
- When we count the number of instances in a category, the result is discrete data
Stevens levels of measurement
- Nominal
- Ordinal
- Interval
- Ratio scale
Nominal scales
- Only categories (labels), not quantities
- Examples: gender, post codes, in health diagnostic criteria, illness or other factor
Ordinal scales
Distinguish objects and events according to their ranking (high or low)
e.g. winner of running race first in speed (finishing first indicates more speed)
Ordinal scales says NOTHING about …?
separation between ranks (difference between first and second place scores)
Interval scales
Compared to ordinal, interval scales add constantly sized (i.e. equal) intervals between all points along the scale
Interval scales hold a true zero or non true zero?
Not true
e.g. 0 in celcius and F do not mean no temperature
Interval scales: Is doubling twice as much?
No. 100 degrees is not as hot as 50 degrees doubled
Ratio scales, does it hold true zero?
Yes
Ratio scales, is doubling REALLY doubling
Yes!
Many physical measures such as length, weight, time, speed, heart rate, and economic measures are measured in _____ scales
ratio
Categorical data…..
- Nominal and ordinal scales because they refer to membership of a category, either a group on a nominal scale (e.g. female) or a rank order position on an ordinal scale (e.g. distinction grade)
- Categorical data are mostly counted, but not quantities
Continuous data…
- Interval and ratio scales
- countable and summary statistics such as averages (can say number of males and females, but not average gender)
Discrete data…(count data)…?
- Counts the numbers of persons, places, things or events
- Expressed only as whole numbers, never fractions
- Number of people in a room, positive test results, births, marriages, deaths
- Because discrete data have equal intervals and true zero, can be analysed as ratio