Evaluation of the Impact of the Implementation of Teaching Scotland's Future

The evaluation offers an overview of the current landscape of teacher education, highlighting what progress has been made in key areas since TSF was published and where further progress and improvements are still needed.

Appendix E: Using logistic regression to examine predictors of aspects of teacher education

14.13 Logistic regression is commonly used to predict the probability of an outcome using several potential predictor variables. In this report, logistic regression was used to model three outcomes:

  • whether Initial Teacher Education (ITE) at university was effective or not in preparing teachers for their first post
  • whether a respondent thought that the professional learning they undertook during their probationary period was effective in helping them to achieve the GTCS Standard for Full Registration
  • whether teachers feel confident that they have the skills they need for their current role.

14.14 For each model, a broad set of potential predictor variables was considered. These variables were explored to see if they helped to explain the outcome. The final predictor variables were selected on the basis that they were either statistically significant or had sufficient explanatory power.

14.15 For all three models, the following demographic variables were included in the initial analysis:

  • sex
  • sector (primary, secondary etc.)
  • full-time vs. part-time
  • permanent vs. temporary employment
  • rurality of school
  • school type (independent vs. local authority).

14.16 For the ITE model, further potential variables included were satisfaction with individual aspects of the ITE course[35]; ITE provider; effectiveness of support provided by the school during their school placement; and support provided by the university.

14.17 For the probation model, further potential variables included were the effectiveness of individual aspects of professional learning during their probationary year[36]; whether the respondent had stayed in touch with university staff; and awareness of the LA/university partnership.

14.18 For the confidence model, further potential variables included forms of professional learning undertaken[37]; whether they agreed that they had a plan for career-long professional learning; whether PRD felt like an ongoing process.

14.19 Variables that were not significant were then excluded and the final models were run. For example, sex was not a significant predictor of confidence in skills and was not included in the final reported model.

Table E.1: Extract of output from the logistic regression confidence model: whether teachers feel confident that they have the skills they need for their current role

B S.E. Wald Sig. Exp(B)
Sector (vs. primary) 14.772 .002
Secondary -.323 .105 9.432 .002 1.382
Special -.249 .201 1.533 .216 1.283
Other -.624 .207 9.136 .003 1.867
Whether have developed a plan for career-long professional learning (vs. strongly agree) 49.328 .000
Agree .399 .174 5.293 .021 .671
Neither agree nor disagree .955 .182 27.636 .000 .385
Disagree 1.049 .189 30.848 .000 .350
Strongly disagree 1.041 .231 20.400 .000 .353
Whether PRD feels like an ongoing process (vs. strongly agree) 17.974 .001
Agree .196 .153 1.629 .202 .822
Neither agree nor disagree .407 .183 4.979 .026 .665
Disagree .520 .172 9.165 .002 .594
Strongly disagree .742 .203 13.325 .000 .476

14.20 The main benefit of using logistic regression in this context is to clearly distinguish the different effects of the various factors. By including in the logistic regression models, for example, both the school sector (primary/secondary) and the school type (LA/independent), it is possible to separate the effect of each of these. This helps overcome the danger that the effect of one variable is confused with the effect of another.

14.21 Three columns in the table above are particularly informative. The first column indicates the different predictor factors included in the model. These can include 'binary' variables (either/or variables), continuous variables (variables that are measured numerically), and categorical factors (variables including a number of different categories). The logistic regression model shows whether each factor has a significant effect on the likelihood of teachers feeling confident that they have the skills they need to do their job once all other factors in the model are controlled for.

14.22 The fifth column, headed 'Sig.', shows whether the factor is significant. A value of less than 0.05 in this column suggests that this factor is significant. So, as the figure for whether a teacher has developed a plan for career-long professional learning (strongly agree vs. strongly disagree) is less than 0.05, it follows that, after controlling for the effect of all other factors in the model, the likelihood of those agreeing they have a plan for their career-long professional learning being confident in their skills is different from those who do not have a plan.

14.23 The second column, headed 'Beta' indicates the direction of the effect. A positive value indicates that those in the first category, for example those with a career-long professional learning plan, are more likely to feel confident, than those who do not.


Email: James Niven

Back to top