Hair et al. (2006: 54-55) note that there are no hard-and-fast rules for
determining exactly how few missing data are acceptable before moving on to
complete the data set with replacement values. General rules of thumb provided
by Hair et al. (2006: 55-56) are that: (A) it should be possible to undertake the
planned analysis effectively using only the cases with no missing values at all; (B)
variables with 15 per cent of values missing are candidates for deletion; (C)
generally, missing data under 10 per cent for an individual case can be ignored,
provided it occurs in a non-random fashion; and (D) removal of a variables or
case is justified if this removal significantly decreases the volume of missing data.
For the remaining 46 item statement variables (53 item statements in total minus
the 7 measuring SQ = 46) across the 224 cases, analyses were performed to
establish: (i) the percentage of variables containing missing vales; (ii) how the
missing values are distributed amongst cases; and (iii) the overall level of missing
data in the data set. Having measured these aspects of the data set, Hair et al.
(2006: 55) recommend further examination to identify non-random patterns of
missing data such as concentration of missing data on specific sets of
questions/statements and attrition in questionnaire completion.
SPSS’s Missing Value Analysis (MVA) module produces summary descriptive
statistics for the proportion of cases and variables with missing values, the
proportion of missing values in each variable and the proportion of all data points
that do not contain a value. The MVA module does not, however, provide a value
for the number of missing values for each case. Accordingly, a new variable
(named Miss_Per_Case) was manually computed in SPSS to provide a sum of
missing values per case.
The MVA analysis reveals that 6 of the 46 variables (13 per cent) in the data set
have no missing values and that 99 per cent of data points are complete.
With regard to rule of thumb (A), the 177 cases with no missing data do provide
an adequate sample size to perform the planned analysis (see the discussion on
sample sizes for SEM in section 6.5.2).
Considering rule of thumb (B), all variables are significantly below the level of 15
per cent missing values that indicates they are candidates for deletion. The
largest proportion of missing values per variable is 3.1 per cent (applying to 4
variables).