Problems in Data Handling

Question 1

Mistreatment of nominal data

Answer 1

when you have nominal data with more than two values don’t compute any statistics other than frequencies

Question 2

what is an exception to treatment nominal data

Answer 2

• if there are only two values then you can do statistics (finding d or r)
• e.g., participant sex: Male = 0, Female = 1

Question 3

How to spot violations of ordinality

Answer 3

if questionnaire items have response options in the form 1= agree, 2= disagree, 3= neutral than the individual scores will not be meaningful (since neutral is not in between agree and disagree)

Question 4

How could you fix violations of ordinality?

Answer 4

recoding the response
e.g., 1= agree, 2= neutral, 3= disagree

Question 5

Out of range values

Answer 5

• when typing questionnaire responses, you might accidentally hit a number that is out of the scale
  e.g., for a 1-7 scale you might hit a 0 or 8 or 44….
• when a missing value is not declared as such; if you do not declare the value (e.g., 9 on a 1-5 scalee) as a missing value then it will be treated as a real response
• both of these examples will distort all statistics

Question 6

Reverse-scored items

Answer 6

• some questionnaires have reversed scales (higher response means less etc.)
• ensure that scale actually measures differences in intended trait not differences in tendency to choose higher response values
• responses to these items must be reversed before calculating scores

Question 7

When reversed score items are not reversed

Answer 7

• if you forget to reverse those items the scale or reverse all items instead of just the reversed ones you won’t measure your trait
• be careful about recoding into same variable; typically better to recode into different variable

Question 8

Data points in the wrong place

Answer 8

• If you do not leave a blank space or add a missing value code for a participants missing data their data for the next variable will get entered for the missing one
• variables are all shifted to the left (dataset becomes meaningless)

Question 9

Switched places for variables

Answer 9

• if two items (or more) are switched when listing items, the stats for these items will be switched
• if these switched items are recoded differently or are scored on different scales it will affect the scale’s properties

Question 10

If you have a set of participants with one set of variables in one file and another set of variables in another file and you want to merge files to analyze the variables together what do you have to do to ensure the data is not mismatched (mismatching of participants with their data)

Answer 10

• make sure that each participant’s data from one file gets merged with the SAME participants data from the other file
  you can do this by:
• sorting the participants identically in both files (all of them in both files)
• think about proper scoring and interpretation of variables
• check files and results carefully always
• failure to do this will result in meaningless data