Missings : what can you do ?

• Listwise deletion (delete rows with missings)
• Pairwise deltion
• Replace by mean

Reliability analysis

• Do the questions relating to the same concept actually measure the same thing ?
• All questions interval/ratio data type :
• Cronbach's alpha
• One general measure (based on correlations) for how well the items are related to each other
• 0 ? ? ? 1

null

• Alpha > .6 = Reliability sufficient : create sum variable
• Alpha <.6 = Reliability not sufficient :
• Could some items be dropped ?
• Separate analyses for X1; X2 etc

Summing items

• Nominal data :
• Gender : 1;= male ; 2 = female
• Answer : 1 = yes; 2 = no
• Can you sum these items ? NO
• Ordinal data :
• Answer 1 : 1= very vad; 2 = reasonable
• Answer 2 : 1 = do not agree; 2 = disagree
• Can you sum these items ? NO
• Interval data (or quasi interval)
• Answer 1 : 1= absolutely not nice; 5 = very nice
• Answer 2 : 1= absolutely not friendly; 7 = very friendly
• Can you sum these items ? PERHAPS
• If cronbach's alpha is larger than .6
• AND
• If items are standardized first : (X - µ)/??

null

Interval data (or quasi interval)

• Answer 1 : 1= absolutely not nice; 5 = very nice
• Answer 2 : 1= absolutely not friendly; 5 = very friendly
• Can you sum these items ? Yes is crobach’s alpha is larger than .6