that there is no association between 3 star / 4 star rating and participation in the
survey (χ^2 p = 0.868; Fisher’s exact p = 0.565). The second alternative test
combined the four and five star hotels into one category and also resulted in a
finding that there is no association between hotels with a 3 star rating and hotels
with 4/5 star ratings and participation in the survey (χ^2 p = 0.823; Fisher’s exact p
= 0.540).
Hotel size was measured based on the number of survey forms sent to each – this
equates with the number of non-supervisory waiting staff employed by each. The
mean number of survey forms distributed to participating hotels was 22.9; for
those hotels that did not return any forms (and therefore effectively did not
participate in the survey) the mean number of survey forms distributed was 19.6.
An ANOVA^10 test was performed to assess whether or not number of non-
supervisory waiting staff had any effect on the propensity of hotels to participate.
This test indicated that there is no association between staff numbers and
propensity to participate in the survey (F = 0.536; d.f. = 1; p = 0.467).
A common reason cited for non-participation was “we are too busy”, although it
was clear from the discussions with the general managers that those who did
participate were more open to this type of co-operation and more positive with
regard to the role of research in hospitality management and organisational
management. There is no reason to believe that the participating hotels had
different approaches to leadership than non-participating hotels.
Regarding the individual respondents, of the 234 responses, a total of 213
useable cases were retained following the data screening process described in
Section 6.6 below. The suitability of this sample size is assessed in two ways:
firstly, it is considered alongside various published guidelines relating to sample
size for SEM analysis; and, secondly, it can be compared with reported sample
sizes for other leadership-focused hospitality studies that have utilised structural
equation modelling.
Recommendations regarding minimum sample sizes for SEM analysis include:
Loehlin (1992) who recommends at least 100 cases, preferably 200
10
NB - the ANOVA test can be used in the same way as the t-test to compare means
between two groups. Hair et al. (2006: 387) describe the t-test as ‘...a special case of
ANOVA for two groups or levels of a treatment variable’.