Hardware Reference
In-Depth Information
Kaiser-Meyer-Oklin (KMO) index
3
and the Bartlett's Test of Sphericity are checked.
Then, principal components as extraction method with Varimax (with Kaiser normali-
zation) as rotation method and the breaks-in-eigenvalues criterion [25] is used to
decide the initial number of factors to keep. Factor loadings equal to or greater than
0.5 are considered strong [25]. Items with low loadings on all components (the cut-off
value for loadings was 0.32 [26]) are eliminated too. Table 7 shows the KMO and
Bartlett´s Test and the extracted components with their loadings. KMO was clear with
value greater than 0.80 and Bartlett´s Test indicates a meaningful relationship among
the variables. The extracted components have been labeled as follows:
Q3.4 Q3.6 Q4.1 Q4.2 Q6.1 Q6.2 = C1 (Market and attitude toward testing)
Q3.1 Q3.2 Q3.3 = C2 (Education and training)
Q5.3 Q5.4 Q5.5 Q5.6= C3 (Integration with development)
Q2.1 Q2.2 Q2.3 Q2.5 = C4 (Career)
Q1.2 Q2.4 Q5.1 Q5.2 = C5 (Attractiveness)
Note that EFA has extracted one factor less than the initial set of the questionnaire:
initial grouping was done by the expert in charge of coordination based on his own
experience and was not object of debate but it was confirmed by experts.
4.2.4 Confirmatory Factorial Analysis
The predictive validation obtained after applying EFA in previous section should be
confirmed to obtain the final model. Confirmatory Factor Analysis (CFA) through
Structural Equation Model (SEM) and Maximum Likelihood (ML) estimation method
is used in order to assess the validity of the model. Several indicators were used to
assess model fit in order to compare the alternative models such as the Root Square
Error of Approximation (RMSEA), Comparative Fit Index (CFI), the Normed Fit
Index (NFI), the Non-Normed Fit Index (NNFI) and the Relative Fit Index (RFI).
Table 8.
Goodness of Fit indicators for the model
Suggested cut-off Values
Factor´s questionarie
χ
2
(df)
189,036 (179)
S- χ
2
>1, <2
1.05
RMSEA
< 0.08
0.0021
CFI
> 0.9
0.997
NFI
> 0.9
0.954
NNFI
> 0.9
0.997
RFI
> 0.9
0.946
Table 8 shows minimum recommended values for good fit [28] as well as the cal-
culated values for the model. All the indicators exceed the minimum recommended
values for good fit providing evidence of discriminate validity.
5 Conclusions
One of the usual shortcomings of the area of software engineering is the lack of
trustable data about which is the state of practice in general and more specifically in
3
KMO: above 0.5, it should be accepted, 0.7-0.8, good value; above 0.8, meritorious [27].
