Scales of Measurement
Assigning numbers to events or objects systematically is termed as measurement, and there are four scales or levels of measurement used: ordinal, nominal, ratio and interval. Nominal measurements make qualitative or categorical grouping and major statistic is the mode. In ordinal scales low numbers represent low values whereas, high assigned numbers represent high value, but intervals between assigned numbers may not be equal. The scale lacks a true zero, for example measurements of attitudes. Interval scales have unit sub-divisions where each unit represents a similar magnitude of the trait assigned to the scale. Ratio scales are akin to interval scales, but have a zero point value point (Sammis, 1997).
An example of a hypothetical research that would make use of the four scales of measurement is the collection statistics on a site so as to assess its performance. A music selling sites that depends on the number of downloads made may conduct this hypothetical research to determine its performance. The statistics to be measured may include the number of hits in a day (ratio scale-they could be zero), the number of visitors from a specific location (nominal scale-categorical in nature), the order in which specific items have received downloads based on quantity (ordinal e.g. 1st, 2nd and 3rd, where the item ranking third has received the least downloads) and finally; the level of satisfaction of the site visitors (Interval-measured on a likert scale- such as satisfied, partially dissatisfied, totally dissatisfied) (Sammis, 1997).
The scale of measure on the number hits shows how much traffic is directed to the site and this is can be a statistical indicator of popularity of the site. The nominal scale measuring the number of visitors from a specific location helps in determining the market location or where there is a certain preference for the site. The ordinal scale on the highest downloaded items in an ordered sequence shows the most popular items of download which can be used to evaluate preference and choice of a category. The appropriate statistic here would be mode. Finally, the interval scale measure on satisfaction measures relative contentment with the site’s services. All these have a high ecological validity because the statistics should be gathered from an actual running site. Reliability improvement can be done through doing re-tests that will compare results from various tests. On the other hand, validity can be improved by clearly defining the objectives and goals of the research (Wren & Phelan, 2010).
Sammis, F. (1997),. Measurements, Benchmark Books
Wren J. and Phelan, C. (2010),. Exploring reliability in academic assessments, retrieved on 19th February 2011 from http://www.uni.edu/chfasoa/reliabilityandvalidity.htm