4. Methodology and Methods
4.9. Major Concerns about Data Handling and Analysis
4.9.1. Quantitative Data
It is difficult to say if there is a relationship between the degree of a teacher‟s job
satisfaction and his or her return of the questionnaire. However, the assumption that those who experienced greater job satisfaction were more likely to return the questionnaire seems less feasible in this study, since teachers‟ comments in the open question session often indicated frustration in their descriptions of challenging conditions and desire for improvement.
The teachers seemed to employ the questionnaire to convey their opinions to me. Thus, it may be that those who were frustrated or less satisfied tended to have responded. Whatever
the case, the survey was not designed to assess teachers‟ absolute job satisfaction but to provide further understanding of relative values in terms of different aspects of their lives (e.g. teachers are less satisfied with their physical environment than their relationship with the community). Therefore, problems with the results of the survey provoked by missing data, or, in this case, no response at all, may be less significant.
In spite of the high return rate of the questionnaire (64.3%), there were data missing from some of them due to the fact that it was lengthy and self-administered. For example, some questionnaires that were returned with a lot of blanks and/or simply ticked with the same response across whole sections were excluded from the analysis, not being considered valid. Organisation of the data and descriptive exploration at univariate and multivariate levels was conducted before any statistical analysis.
Missing data is handled in this study in two ways. On one hand, as one of my intentions is to get a picture of the two districts by using quantitative data, I use the data from all 847 cases for descriptive analysis, indicating missing frequencies for each variable. On the other hand, listwise deletion – that is, the exclusion from analysis of cases with missing data – is employed for factor analysis in order to explore teachers‟ perception and satisfaction. The sample size is thus reduced to 416, which, however, is large enough for factor analysis (Oppenheim 1992; Allison 2002).
The 416 responses indicate that on average, 38.7% of teachers are trained: 19.0% in Ponkujaku and 48.9% in Aumisoe. These findings are consistent with statistics from the 2007/08 school census of each DEO, which indicates that on average, 36.0% of a total 1,318 basic school teachers are trained: 22.2% in Ponkujaku and 41.8% in Aumisoe. This shows that the survey sample has little selection bias in terms of district and teachers‟ qualifications. Similarly, the sample reflects the general population in terms of male and female teachers, 84.6% and 15.4% of the sample being male and female respectively, while 84.2% and 15.8% of teachers are male and female respectively in the general population. Therefore, the survey sample used for factor analysis fairly represents the populations of the case study districts.
In the questionnaire, some items had slightly less responded to than others. Although there is no missing data in respect of district, gender or qualifications (trained and untrained), of 847 responses, 36 (4.3%) did not include their age. All questions on teachers‟ perceptions and levels of job satisfaction also had some missing data, which ranged from 29 (3.4%) to 65 (7.7%).
It seems that teachers were somewhat reluctant to answer questions about their colleagues; 7.7% and 6.6% of all responses lack data on satisfaction with their head teachers and colleagues respectively. They may have thought that they could be identified as the questionnaires were returned to me by the schools, although each individual response was anonymous. Other questions with high levels of missing data – indeed, the four highest – are those concerning how the media projects teachers (7.7%), safety (7.3%), retirement benefits (7.0%) and the health service teachers are entitled to (6.7%).
Teachers may have left items blank because they regarded those issues to be irrelevant or inapplicable; additionally, asking the age of a teacher might have been a sensitive area. However, the rate of missing data is no higher among those who did not include their ages than among those who did, with a differential ranging from −0.2% to 0.3% for each perception and satisfaction question. In conclusion, the rate of missing data – in other words, selection bias – does not seem to be a serious problem.
4.9.2. Qualitative Data
Interpreters were necessary in order to interview most community members, unlike basic school teachers, who spoke English. I planned to employ somebody from outside the community and recruited an English/Twi (local language), tutor at a teacher training college as an interpreter. However, I found that he could not help sharing his opinions with those I was interviewing. I also felt that he tended to summarise accounts. His use of complex phrases made me less convinced of how genuinely he was interpreting. Following this experience and with other logistical difficulties, such as transportation and
accommodation for interpreters from outside, I sought senior secondary school graduates in the community. I was much more comfortable with them, as they attempted to interpret faithfully without skimming over the details. Nevertheless, I was always uncertain that
what was rendered into English was the same as the community member had said. However, finding people who spoke English also implied a selection bias.
In this study, evidence was collected with the use of various methods during field research that lasted a year. Deciding how I was to incorporate the various data – including that derived from the lower level school community cases – in order to write up my „case‟ was a significant challenge.
As this was a multiple-case study, there were two methods of analysis, that is, the „within- case‟ study analysis and the „cross-case‟ analysis. Merriam (1998) suggests two stages: “once the analysis of each case is completed, cross-case analysis begins” (p195 emphasis in the original), citing Miles and Huberman (1994):
Cross-case analysis is tricky. Simply summarizing superficially across some themes or main variables by itself tells us little. We have to look carefully at the complex configuration of processes within each case, understand the local dynamics, before we can begin to see patterning of variables that transcends particular cases (p 205– 206).
This thesis follows the above procedure in order to report the „cases‟, in the belief that such a method is necessary to obtain a better understanding of the highly complex nature of teacher motivation in any given context.