levels of measurement

• factors affecting the choice of stat test

  • if an appropriate test isn't selected and justified, otherwise the statistical analysis may be brought into question
  • the different factors are:
  • experimental design
  • levels of measurement
  • nominal data
  • ordinal data
  • interval data
• difference or relationship

  • must decide whether the hypothesis is looking to investigate a difference or relationship first as most stat tests are designed to be used for one specifically
  • differences - 2 conditions, one control and one experiment group
  • relationships - seeing if one variable impacted another
• experimental design

  • secondly, we must identify which research design was used
  • only considered when looking for a difference
  • experimental design will be one of the following 3:
  • independent groups
  • repeated measures
  • matched pairs
  • from this the type of data (related or unrelated) can be decided
  • related- participants are related in some way (matched pairs or repeated measures)
  • unrelated- 2 separate groups in each condition (independent groups)
• levels of measurement

  quantitative data falls into one of 3 levels of measurement

  -nominal

  -ordinal

  -interval
• nominal

  • categorical data e.g. wanting to know if people went to a school or college when as they're separate categories
  • each ppt will only appear in one category which is known as discrete data
• ordinal

  • data ordered in some way and the intervals between the data aren't equal
  • used to rank data where the data is only used to state where it ranks in relation to other scores
• interval

  • data that is ordered (like ordinal) but the intervals are equal between each measurement
  • much more objective and scientific in nature
  • e.g. time and temperature (difference between 3 and 4 degrees is the same as between 35 and 36 degrees
• evaluating nominal data

  +generated quickly from closed questions or interviews scan be tested quickly for reliability

  -data can appear too simplistic meaning there's no scale of referance
• evaluating ordinal data

  + provides more detail than nominal as the scores are ordered in a linear fashion (highest to lowest)

  -intervals or scores aren't equal meaning the mean can't be used as a measure of central tendency. however, the median is used to overcome this
• evaluating interval data

  + more informative than ordinal and nominal as the intervals are of equal value/distance meaning they're more reliable

  -sometimes intervals can be based on random choice