Summary
Highlights
This lecture, the last for postgraduate students in the Faculty of Education, focuses on correlation coefficients. Previous lectures covered comparison tests like ANOVA and t-tests. Correlation coefficients are widely used in social sciences, education, psychology, and even medicine. The video will specifically address the simple bivariate correlation coefficient, which examines the relationship between two variables. Other types, such as multiple correlation (between one variable and a set of others) and canonical correlation (between two sets of variables), are mentioned as available in SPSS but will not be covered in this course.
A bivariate simple correlation is used to find the relationship between two variables. It's 'simple' because it assumes a linear relationship between the variables, meaning their relationship can be represented by a straight line. While non-linear, curvilinear relationships exist, they are beyond the scope of this course. A correlation coefficient's value ranges from -1 to 1. The sign (positive or negative) indicates the direction of the relationship (positive for direct, negative for inverse), and the absolute value indicates its strength. A positive relationship means both variables increase or decrease together, while a negative relationship means one increases as the other decreases.
The video presents an example with variables like workplace bullying, burnout, depression, and job satisfaction. It's expected that burnout, bullying, and violence in the workplace would have positive correlations with each other, meaning an increase in one leads to an increase in the others. However, job satisfaction, being a positive attribute, is likely to have a negative correlation with negative attributes like burnout and depression. The importance of understanding variable types (nominal, ordinal, scale) and ensuring they are correctly configured in SPSS is also highlighted, with an example of changing 'educational level' to ordinal.
To calculate the bivariate correlation in SPSS, navigate to 'Analyze', then 'Correlate', and select 'Bivariate'. The video demonstrates how to input variables, such as burnout and depression, into the analysis. It highlights that SPSS offers different correlation methods: Pearson (for linear, normally distributed data), Spearman (for ordinal data or non-normal distributions), and Kendall's tau-b (for ordinal data with tied ranks). For this session, Pearson's correlation is selected. The concept of one-tailed vs. two-tailed tests is also mentioned, where one-tailed is used if the direction of the relationship is hypothesized, and two-tailed if it's not. The 'flag significant correlations' option is generally discouraged in academic research.
The SPSS output for correlation provides three key pieces of information for each pair of variables: the Pearson correlation coefficient, its significance level (p-value), and the sample size. A correlation of 1 signifies a perfect positive correlation (a variable with itself). For burnout and depression, a coefficient of 0.605 indicates a strong positive relationship, meaning increased burnout is associated with increased depression. The p-value, often displayed as '0.000', means it's less than 0.001, indicating high statistical significance. The video criticizes the use of asterisks (one for p<0.05, two for p<0.01) in academic papers, advising to report the exact p-value or 'p < 0.001' instead.
When calculating multiple correlations, the sample size (N) might vary for different variable pairs. This is because SPSS, by default, uses 'pairwise deletion' for missing data. This means if a participant has missing data for one variable in a pair, they are excluded only from that specific correlation calculation. An alternative, 'listwise deletion', excludes any participant with missing data on ANY of the analyzed variables, leading to a consistent but potentially smaller sample size. While listwise deletion provides a uniform sample size, it might lead to a significant loss of data if many cases have scattered missing values. For large datasets with minimal missing data, listwise deletion might be acceptable.
The video revisits the direction and strength of correlations. Job satisfaction is highlighted as the only positive variable, contrasting with negative ones like burnout and depression. Therefore, job satisfaction is expected to have negative correlations with the other variables, which is confirmed in the output. For example, increased burnout is associated with decreased job satisfaction. The strength of the correlation (e.g., 0.6 for burnout and depression vs. 0.3 for burnout and workplace violence) indicates how closely variables are related, with values closer to 1 (or -1) indicating stronger relationships, and values closer to 0 indicating weaker relationships. Exact thresholds for 'weak,' 'moderate,' and 'strong' are not universally agreed upon but generally, values above 0.5 or 0.6 are considered moderate to strong.
To compare correlation coefficients between different groups (e.g., males and females), the 'Split File' function in SPSS can be used ('Data' -> 'Split File' -> 'Compare groups' and select 'Gender'). This will generate separate correlation tables for each group. The example shows that correlations between negative psychological aspects (like depression and burnout) are higher for females than for males, suggesting females might be more affected by these negative factors in the workplace. This allows for interesting comparative analysis.