Video 1 – week 2
SERVQUAL (Parasuraman et al., 1988; Cronin and Taylor 1992)
22-item instrument was developed. Exploratory factor analysis confirmed
5 dimensions summarizing the 22 items:
1. Tangibles (4 items: T1, T2, T3 and T4)
2. Reliability (5 items: RL1, RL2, RL3, RL4 and RL5)
3. Responsiveness (4 items: RS1, RS2, RS3, RS4 and RS5)
4. Assurance (4 items: A1, A2, A3 and A4)
5. Empathy (5 items: E1, E2, E3, E4 and E5)
- The questions are called items in the exploratory factor analysis. So
for tangibles there are asked 4 questions.
Application of SERVQUAL to the Mensa
- Research objective: validate the SERVQUAL scale in this new
setting and assess its reliability.
- Requires testing the dimensionality of SERVQUAL Sale: Is it 5
dimensional in this new setting?
- Survey, N = 209
Task
1. Assess the dimensionality of SERVQUAL Scale: past research
suggests a 5-dimensional structure. Is the structure of the scale in
this new setting consistent with previous research?
a. This requires two key interpretations in the resulting factor
solution
- The number of dimensions (factors extracted) and
- How variables ‘clump/group together'
, 2. Asses the reliability of the scale
3. Create summated scales for further analysis
Steps for the factor analysis
Task 1: Dimensionality of SERVQUAL
1. Factor analysis is conducted via: Analyze > dimension reduction >
factor
2. Select 22 variables (items on the scale). Move them to ‘variables’
3. Under ‘descriptives’ select ‘KMO and Bartlett’s test of sphericity’
4. Under ‘extraction’ make sure ‘principle components’ is shown in the
methods sections and select ‘scree plot’ in the display section. In the
extract section, select ‘Based on Eigenvalue’ (or if you want to force
a specific number of factors, click on ‘fixed number of factors’ and
type the number)
5. Under ‘Rotation’ select ‘Varimax’
,KMO and Barlett’s test
- What does it mean? This tells whether it is appropriate to conduct a
factor analysis in the data set. So the correlation should be
significant below 0.05 (in this case 0.00). The upper number should
be above 0.05 (in this case 0.73)
Total variance explained
- We detect how many factors come out from the Eigenvalues higher
than 1 rule. In this case there are 7 factors higher than 1 and these
7 factors together conducts 62% of variance that is in the original 22
items.
- This table tells us how much of the variance (information) in the
original variables is captured by each of the factors (e.g how much
variance would the first factor capture by itself..)
- The number of components (factors) equals the number of variables.
(In this case 22) However, we see that later/lower components add
little (don’t capture much more information) By default, SPSS will
only continue with those components with an EV > 1 (in this case 7
factors)
o Rationale: any individual factor should account for the
variance of at least a single variable if it is to be retained!
o Eigenvalue: the amount variance that is explained by a factor
(= column sum of squared factor loadings)
, Scree plot
- The elbow rule: the sudden flatting in eigenvalues. In this case it is
at 8 and we always do minus 1 so it is 7. (the same as we could
conduct from the total variance explained table)
So is it consistent with the prior research? No in the beginning we thought
there were 5 factors (tangibles, reliability, responsiveness, assurance and
empathy) and now we see there are 7 factors.
Communalities
- Tells us how much of the variance of the original variable is captured
by the extracted factors
- If the extraction is low <0.3 it means it correlates weakly with the
other items. It is unique and measures something else, these items
should be excluded.
Rotated component matrix
SERVQUAL (Parasuraman et al., 1988; Cronin and Taylor 1992)
22-item instrument was developed. Exploratory factor analysis confirmed
5 dimensions summarizing the 22 items:
1. Tangibles (4 items: T1, T2, T3 and T4)
2. Reliability (5 items: RL1, RL2, RL3, RL4 and RL5)
3. Responsiveness (4 items: RS1, RS2, RS3, RS4 and RS5)
4. Assurance (4 items: A1, A2, A3 and A4)
5. Empathy (5 items: E1, E2, E3, E4 and E5)
- The questions are called items in the exploratory factor analysis. So
for tangibles there are asked 4 questions.
Application of SERVQUAL to the Mensa
- Research objective: validate the SERVQUAL scale in this new
setting and assess its reliability.
- Requires testing the dimensionality of SERVQUAL Sale: Is it 5
dimensional in this new setting?
- Survey, N = 209
Task
1. Assess the dimensionality of SERVQUAL Scale: past research
suggests a 5-dimensional structure. Is the structure of the scale in
this new setting consistent with previous research?
a. This requires two key interpretations in the resulting factor
solution
- The number of dimensions (factors extracted) and
- How variables ‘clump/group together'
, 2. Asses the reliability of the scale
3. Create summated scales for further analysis
Steps for the factor analysis
Task 1: Dimensionality of SERVQUAL
1. Factor analysis is conducted via: Analyze > dimension reduction >
factor
2. Select 22 variables (items on the scale). Move them to ‘variables’
3. Under ‘descriptives’ select ‘KMO and Bartlett’s test of sphericity’
4. Under ‘extraction’ make sure ‘principle components’ is shown in the
methods sections and select ‘scree plot’ in the display section. In the
extract section, select ‘Based on Eigenvalue’ (or if you want to force
a specific number of factors, click on ‘fixed number of factors’ and
type the number)
5. Under ‘Rotation’ select ‘Varimax’
,KMO and Barlett’s test
- What does it mean? This tells whether it is appropriate to conduct a
factor analysis in the data set. So the correlation should be
significant below 0.05 (in this case 0.00). The upper number should
be above 0.05 (in this case 0.73)
Total variance explained
- We detect how many factors come out from the Eigenvalues higher
than 1 rule. In this case there are 7 factors higher than 1 and these
7 factors together conducts 62% of variance that is in the original 22
items.
- This table tells us how much of the variance (information) in the
original variables is captured by each of the factors (e.g how much
variance would the first factor capture by itself..)
- The number of components (factors) equals the number of variables.
(In this case 22) However, we see that later/lower components add
little (don’t capture much more information) By default, SPSS will
only continue with those components with an EV > 1 (in this case 7
factors)
o Rationale: any individual factor should account for the
variance of at least a single variable if it is to be retained!
o Eigenvalue: the amount variance that is explained by a factor
(= column sum of squared factor loadings)
, Scree plot
- The elbow rule: the sudden flatting in eigenvalues. In this case it is
at 8 and we always do minus 1 so it is 7. (the same as we could
conduct from the total variance explained table)
So is it consistent with the prior research? No in the beginning we thought
there were 5 factors (tangibles, reliability, responsiveness, assurance and
empathy) and now we see there are 7 factors.
Communalities
- Tells us how much of the variance of the original variable is captured
by the extracted factors
- If the extraction is low <0.3 it means it correlates weakly with the
other items. It is unique and measures something else, these items
should be excluded.
Rotated component matrix
