I am analysing the data of a cross-sectional study I am working on. It explores differences in self-compassion, burnout and skills between clinicians joining a peer supervision group and clinicians not joining this group but receiving other forms of supervision.
In the correlation table, the variable 'participants to peer consultation' has no significant correlation with anything. Other variables correlates and don't show this anomality. It is only this one that is a bit funky. Any ideas why? I put an example here below.
Pearson .052
Sig. (2-tailed) .602
N 103
Thanks!
The reasons why a correlation coefficient is not statistically significant are several, and can include:
1. There is no linear relationship between the variables involved. However, the relationship may be of another nature (eg polynomial relationships)
2. There are outliers that cause the relationship to be non-significant
Therefore, it would be advisable to carry out analyzes to detect possible influential outliers, as well as to study non-linear relationships between the variables.
It means that increase or decrease in one variable is not linked to an increase or decrease in the other variable. There is not a linear correlation
Correlation shows the relationship between associated variables, for example, X and Y, however, in your case the nonsignificant mean X is not associated with Y, either positively or negatively. moreover, if you want to check the change in Y through X for this you have to run regression analysis.
In linear regression, if we estimate a model with a dependent variable (Y) and a single independent variable (X), the estimated coefficient for the standardized solution coincides with Pearson's correlation coefficient between X and Y. In this case it is assumed that the relationship between X and Y is linear. However, we can consider other types of nonlinear relationships in the form of a U (inverted or not), or even with an S shape.
For example, to analyze U-shaped relationships between X and Y we would estimate the following model:
Y = a + b1 X + b2 X ^ 2 + e
Where X^2 represents X squared (it is recommended to center X before calculating the square and estimating the regression)
I recommend to see:
Aiken, L.S. y West, S.G. (1991): Multiple Regression: testing and interpreting interactions, Sage Publications, Thousan Oaks, CA
https://t.ly/snN9
https://t.ly/7LeS
https://www.econstor.eu/bitstream/10419/63088/1/547244843.pdf
https://statisticsbyjim.com/regression/choose-linear-nonlinear-regression/
It sounds like your variable of interest is dichotomous. Consider point-bisexual correlation or perhaps a t-test. Do a Scattergram perhaps to get a visual impression of the data.
Irwin Altrows Sandro Voi
A clarification: Irwin wanted to say "point-biserial" correlation, not "point-bisexual"
https://en.wikipedia.org/wiki/Point-biserial_correlation_coefficient
Simplemente que ambas variables coexisten de manera no significativa, es decir X y Y, son valores que se dan de manera independiente
Weak R2 of 0.15 is likely influenced by factors outside your control. Generally, even when R2 is low, it can be significant, indicating that your regression model has statistically significant explanatory power. In social studies, it is common that many variances are outside the control of the researcher and a low R2 is usually expected. Because it may depend on people’s own interpretations; a lower R2 may add value even though other factors are unknown or not measured.
- Badges
- Science topic
- Similar topics
- Social Psychology
- Attitude