Exploratory Factor Analysis --- Dataset (TOSSE-R.sav). Presenter : Melody Date: June 1, 2013. Suitable for FA? Based on what? Stages of making a decision on the factors to be extracted What is the convergent validity? discriminant validity?
Presenter : Melody
Date: June 1, 2013
These items are eyesores.
Q6 (r = .271), Q7(r = .225), Q10 (r =.254), Q12 (r =.079), Q19 (r = - .095), Q20 (r = .171), Q23 (r = .281), Q25 (r =.176), Q26 (r = .151), and Q27 (r = .259)
Why? The standard that the extent of association among items should be within 0.3~0.8 is not met.
singularity Q12 (factor loading value is 0.297)
indicator of sampling adequacy,
the Measure of sampling adequacy (MSA)
Suitable for FA, but some items had better be crossed out.
an action taken: Q12 (singularity problem) and Q10 (comparatively low factor loading value =0.417< 0.5) deleted.
an action taken : the remaining items (26 items) are under EFA by resorting to abliminrotation approach. ( because of expected correlated underlying factors)
Q21 and Q27 crossing-load on two
the loading values of Q1, Q9, and
Q11 are suppressed due to their
coefficient values below the
threshold set as 0.4.
Q21, Q27, Q1, Q9, and Q11 deleted.
21 items are left for EFA again.
communalities values after extraction > 0.7
( if the # of variables is less than 30 )
sample size > 250
average communality > 0.6
retain all factors with eigenvalues above 1
(no crossing-loadings between factors )
variables precisely loading on factors
Correlations between factors do not exceed 0.7
Overall Reliability of the 21 items in the dataset (TOSSE.sav.)
Larger than 0.7
Reliability of Comp 2
Reliability of Comp 4 =. 0.7
Reliability of Comp 5 > 0.7
Successfully Applying Statistics to
Applying Statistics to Experiment’
Why don’t we first group the question items into four components in correspondence with the four characteristics proposed by Bland, and then run FA? CFA?