Qualitative Data

1
Qualitative Data

• Qualitative data can cover a multitude of
  situations
• Subjective scales
• Textual information
• Photographs etc.
• Sometimes it is used as a cover for sloppy work!
• The CRITICAL issue is what is inferred from the
  data.

2
Subjective Scales

• So called Likert scales 1 5 or 1 7 for
  opinions
• Scales of suffering for animals
• Scales for condition of animals

3
(No Transcript)
4
Subjective Scales

• Results may be Ordered
• BUT
• Differences may not be consistent
• Ratios may not be consistent

5
Other Subjective Data

• Text
• NVivo (Formerly known as NUDIST)
• Other forms of data

6

• The sampling needs to be carefully considered if
  we want to infer anything about a wider
  population
• Beware of Information Rich sampling

7
Contingency Tables

• Often qualitative data is collected into tables
• Tabulate in Minitab

8
Chi Squared Test for Independence

• In a two way table we can calculate expected
  counts assuming that the pattern down the
  columns (or across the rows) is constant

9
Chi Squared Test for Independence

• In a two way table we can calculate expected
  counts assuming that the pattern down the
  columns (or across the rows) is constant

10
Chi Squared Test
11
Sample Sizes

• The one question everyone wants to know is how
  big a sample do I need?
• The answer again breaks down into whether we are
  interested in means or proportions.

12
Sample Size for Means

• I use three methods
• The Scientific Method
• From the Confidence Interval we can show
• This is of limited value as we rarely know s and
  often dont have any idea of m and is only true
  for a single mean
• The Research Method
• Look at similar research and see what sample
  sizes they used Did they get significant
  results?
• The Resources Method
• What does it cost to process one unit?
• How much resource do you have altogether?
• Divide one by the other

13
Sample Sizes for Proportions

• The situation with proportions is a little
  different
• For as single proportion we can write the
  worst case is when p 0.5 then