CHAPTER 5: CONCLUSION
5.4 LIMITATIONS OF THE STUDY
Like in any research effort several limitations have been identified, which might have influenced the results. It is therefore important that readers take cognisance of these limitations and understand the possible impact on the research findings, before any inferences can be concluded.
The first limitation in the current study was the high drop-out rate amongst control, but also experimental group (although to a lesser degree) participants. First and foremost, the dropout rate directly influences the sample size, which has a critical impact on the statistical power of the experiment. According to Howell (2004) sample size is a common variable that affects the power of a test. The main reason being that means, and differences between means (standard deviations) are used to investigate the relationship between constructs, and hence, the sampling distribution becomes an important aspect in the final judgement. The variance of the sampling distribution of the mean decreases when either n increases or σ2 decreases (Howell, 2004). Therefore, the small sample size utilised at T3 could have had a negative
impact on the statistical power of the results, when evaluating the effects of the EI training programme. Secondly, dropout rates provide information regarding participant’s interests in the training programme. Due to the fact that participation in the programme was voluntary and students were free to withdraw at any moment, there were many students who left the programme without prior notice or reason. It can be presumed that these students lost interest, or simply did not have any interest in the programme to start off with. This lack of interest can cause students to rush through assessment sessions and randomly or inconsistently provide answers to measurement questions, ultimately contaminating inferences drawn from the data.
According to Babbie and Mouton (2010) another possible limitation when collecting data for the social sciences is the utilisation of self-report instruments. This method has been criticised for two main reasons (Conway, 2002). Firstly, the inferences made by the researcher regarding the correlations and causal relationships between the variables in question may be artificially inflated as a result of common method variance. Secondly, the data may be prone to response biases. A common response bias is social desirable responding that occurs when respondents try to create a favourable impression of themselves by over-promoting admirable attitudes and behaviours, while under-reporting attitudes and behaviours which they feel are not socially acceptable or respected (Zammuner & Galli, 2005). This study utilised a battery of self-report questionnaires, and the relevant findings should be interpreted in terms of this drawback.
A third limitation was that of confounding variables. A confounding variable is an uncontrolled extraneous variable that co-varies with the experimental manipulation, which undermines the internal validity of the experiment (Terre Blanche, et al., 2006). This could relate to any experience outside of the actual intervention. Specific confounding variables in the current study are situational and time specific variables to which the respondents could have been exposed to at the time of the assessments or training. More specifically the mid- year June exam commenced just after the third assessment period. Students therefore experienced extreme workload pressures and stressors during this timeframe. This could perhaps also explain the high dropout rate from T2 to T3 – especially in the control group where the commitment to the research was probably not as prominent as it was for experimental group participants.
Another limitation perhaps lies within elements related to the facilitation of the training programme. Due to the fact that a relatively large amount of students had to receive training, and the fact that groups had to remain small enough to encourage open discussions (i.e. no more than four individuals per group), trained individuals were needed to help facilitate the EI programme. These facilitators therefore needed to become familiar with the content of the programme prior to the implementation of the programme. Although great efforts were made to empower facilitators with the necessary knowledge and skills to effectively communicate the programme contents to participants, it could be that some facilitators still struggled to effectively transfer the learnt contents of the programme to students. This could be especially so for the more complex EI skills, such as the emotional regulatory components. Beyond the knowledge base of the facilitators their enthusiasm, social skills and EI, and attitude at the time, could also have determined how much buy-in they would have received from participants in their groups. Therefore, the effects of the training programme were heavily dependent on these factors related to of the facilitators, which were unknown elements in the current research.
Some of these above mentioned limitations are familiar to the controlled experimental design and will always be present to some extent, unless the experimental design is altered. However, given the limitations listed above it needs to be pointed out that certain learning points from previous empirical studies were incorporated into this research effort, such as using a control group (and stratifying the control and experimental groups better), as well as adding a secondary post-test (Burger, 2009) to gauge the sustainability if the intervention over time. Furthermore, a clear attempt was made to utilise a larger sample size as was suggested by Herman (2012). In this respect the current research was successful to a certain extent.