Hello! I'm currently doing my Thesis on interpretation biases, loneliness and pain. I want to look at the contribution of interpretation biases to loneliness and then the contribution of interpretation biases to pain separately.
The ''interpretation bias'' scale consists of 4 subscales, namely, a) negative interpretations of social evaluation, b) benign interpretations of social evaluation, c) negative interpretations of bodily harm, d) benign interpretations of bodily harm.
My hypotheses are : 1. Threatening interpretation biases (particularly those indicating bodily harm) predict pain
2. Threatening interpretation biases (particularly those indicating social threat) predict loneliness
I also want to control for Anxiety and Depression since they have been found to be associated with both pain and loneliness as well as interpretation bias.
My supervisor advised me to do a standard multiple regression including all of the interpretation bias subscales as predictors and pain as a criterion. She told me that although we expect to see the 'negative interpretations of bodily harm' item to be significantly associated with pain we will include all of the other items of the scale in the regression because research has shown that for example weaker endorsement of benign inerpretations of bodily harm na be also associated with pain. However, I'm not sure. Can I include all of the items in the regression model and see the individual effect of each of the interpretation bias items on pain or is that wrong?
Then I was told that I need to do a second standard multiple regression including my covariates (anxiety and depression) in order to control for them and see if there is any difference.
Exactly the same process I am expected to repeat for my other dependent variable ''loneliness'' but in that case I expect that negative interpretations of social threat to be significantly associated with loneliness.
Do you think I need to continue with standadrd multiple regression or to use another kind of regression such as hierarchical multiple regression? I don't know if the right method to control for the covariates is just to do another standard multiple regression and included these along with the interpretation bias subscales.
While I'm reading different research papers I see that many authors use hierarchical regression but I'm still not sure which is the right method because I'm thinking that I don't want to find the best model as hierarchical regression does. But still I'm not sure if I need to keep it simple and continue with the standard one?
If you do as suggested by your advisor would be 4 hypothesis 4 effects (2 subscales in the 2 dependent variables), according to your approach to hypothesis suggests two separate analysis with the specific variables, it is also best to use hierarchical regression that fits what you propose considering that part of the study includes control of other variables as well. Although your consultant's suggestion is more possible in SEM path analysis models where even the items of each instrument are added unlike previous analyses where the total scores of each variable are included.
Cristian Ramos-Vera appreciate your suggestions. I was thinking of two ways to build the hierarchical multiple regression. The first one is that I could control for depression and anxiety in the 1st step and then in the 2nd step to include all of the 4 subscales of interpretation biases (negative bodily threat, positive bodily threat, negative social threat, positive social threat) and to repeat the same process for loneliness which is my other dependent variable.
The other way I was thinking about is that since I'm expecting that negative interpretations of bodily threat will predominantly contribute to pain I don't know if it's more appropriate to build a 3-step hierarchical model and enter in the 1st step my covariates (depression, anxiety) in the 2nd step all the other interpretation bias subscales (negative social threat, positive social threat, positive bodily threat) and then in the 3rd step the negative interpretations of bodily threat.
Then I could repeat the same process for loneliness given than I'm expecting that in that case negative interpretations of social threat will mostly contribute to loneliness. So, I was thinking of including my covariates (depression, anxiety) in the 1st step, all the other three items of interpretation bias (negative bodily threat, positive bodily threat, positive social threat) in the 2nd step and then in the 3rd step negative interpretations of social threat.
If I'm gonna compute hierarchical regression which option do you think is better? A 2-step or 3-step regression model for loneliness and pain?
- Badges
- Science topic
- Similar topics
- Philosophy
- Esthetics
- Interpretation