week 4: normality/outliers/erroneous data

first step of data screening is checking for

• erroneous data

• normal distribution

erroneous data eg age results from the methodology/materials:

• time taken

• extreme/non-serious responses

• questionnaire/subscale scores are out of range

ways to screen data:

• check demographic data

• time check - how long did it take them to do?

  • what is reasonable?

• check responses vary

  • have people just ticked strongly agree for everything?

• check response ranges - within expected ranges?

• check for ppts with missing data

removal of ppt needs to be documented and justified

ways to reduce poor/nonsense responses:

• commitment checks

• attention checks eg please tick the neutral response for this statement, trap questions

• speed warning eg you’ve answered these questions very quickly, have you given everything sufficient thought

assumptions = a rule that data needs to meet

parametric tests rely on the mean - 3 assumptions

• interval/ratio data

• homogeneity of variance

• normally distributed

non parametric tests rely on the median

kurtosis = skew of spread about the mean

• platykurtic = flat distribution with thin tails - negative kurtosis

• leptokurtic = fat tails - positive kurtosis

• mesokurtic = 0

value of kurtosis is 0 if data is normally distributed