Appendix B: Sampling approach
14.1 In essence the aim of any sampling exercise is to create a subset of the survey population that is as representative as possible of that population. As such it was important that the quantitative data collection should achieve a representative sample of teacher views. However, as different questions were asked of different staff groups, the survey also needed to be capable of providing robust and representative estimates of teachers' views, disaggregated by sector and staff group (e.g. estimates for special school head teachers, secondary student teachers etc.).
14.2 The sampling frame used was the GTCS teacher register; those with incomplete data were excluded from the sampling procedures. In order to allow subgroup analysis a disproportionate sample stratified by sector and staff group was drawn. This involved sorting and splitting the sample into different groups (e.g. secondary head teachers, primary student teachers etc.), then within each group (or strata) selecting at random a certain number of respondents, based target sample calculations (see Table A.1). This approach ensured that for every subgroup a sufficient number of respondents were selected to allow analysis. In order to allow overall estimates representative of the population, weighting was applied (full details can be found in Appendix C). For a number of groups (supply teachers, early years teachers, those working in FE, those working in HE, those working in central locations, those working in the special sector) the numbers available to sample were low so all possible respondents were selected.
14.3 The sample also had to reflect socio-economic, geographical (both in terms of LA and rurality) and ethnic differences. These variables were not set up as formal strata, as with the sector and staff group, because they did not need to be disproportionately sampled. Rather they needed to reflect the wider population profile. In order to achieve this, respondents within each strata were sorted by local authority, rurality of school, percentage of BME pupils at school, school size and SIMD of school. School level variables were taken from the Scottish Government schools database and matched into the GTCS teacher register.
14.4 Respondents were then selected at random before using a 1 in n, approach to select respondents. The sampling calculation differed within each strata based on the selected sample number required and the total population. So, for example if there was a population of 1,000 and the selected sample number required was 100, the sampling calculation would be 1 in 10.
14.5 In summary, the sampling process was as follows:
- Match school information into the GTCS teacher register
- Stratify the sample by sector and staff group
- Within each strata, sort the respondents by school level variables (rurality, percentage of BME pupils, SIMD, school size) and local authority
- For each strata select every 1 in n respondents, based on selected sample calculations.
14.6 Due to a lower than anticipated response rate, halfway through fieldwork a second sample was drawn to boost responses, the same sampling approach was taken as outlined above.
Table A.1 Target sample numbers
|Primary (assumed 35% response rate)||Total population||Original selected sample||Additional selected sample||Target achieved sample|
|Depute head teacher||1,124||1,124||393|
|Secondary (assumed 25% response rate)|
|Depute head teacher||926||926||232|
|Special (assumed 35% response rate)||1,825||1,825||639|
|Centrally employed (assumed 35% response rate)||972||972||340|
|Supply teachers (assumed 35% response rate||799||799||280|
|Independent schools (assumed response rate 35%)||2,979||1,486||1,493||520|
|Early Years (assumed response rate 35%)||222||222||78|
|Those working in FE (assumed response rate 35%)||121||121||42|
|Those working in HE (assumed response rate 35%)||266||266||93|
Email: James Niven