Hello how could I have the sig. for my part correlations in this case in spss? I have watched many video but seems like no one points out this. Is there no way in SPSS to have sig. for my part correlations? Thanks!
Hi Louisa. Here is a link to a nice article: online.stat.psu.edu/stat505/lesson/6/6.3 The page details how to calculate an appropriate t-statistic for your part correlations, by-hand. You can then look up the critical value on a set of t-tables to get the cut-off point. I hope that helps. Regards. Jonathan
Using SPSS: Unfortunately, I know of no way to get this from the Regression dialog. But what you could do is the following, assuming we are happy with the differences between partial and part correlations: Part Correlation significance: For each independent variable in your regression you have a zero, partial, and part correlation listed. Why not regress that independent variable (which becomes a new dependent variable) on all other independent variables, but save the residuals: Analyse -> Regression -> Linear -> Save -> Unstandardised Residuals. This will create a new variable in your SPSS datafile. That variable effectively represents the unaccounted for variance. Now correlate that new variable with your original dependent variable. Analyse -> Correlation -> Bivariate. The significance listed in that analysis will be associated with the part correlation. You would do the above for each independent variable. If you are interested in the same analysis but for the partial correlations you would need to regress your original dependent variable on all independent variables except the one you are testing the partial correlation significance of. Oh, and save the residuals. The newly saved variable is the dependent variable without the shared variance from the independents. I do hope this helps. Regards. Jonathan.
Hello how could I have the sig. for my part correlations in this case in spss? I have watched many video but seems like no one points out this. Is there no way in SPSS to have sig. for my part correlations? Thanks!
Hi Louisa. Here is a link to a nice article: online.stat.psu.edu/stat505/lesson/6/6.3
The page details how to calculate an appropriate t-statistic for your part correlations, by-hand. You can then look up the critical value on a set of t-tables to get the cut-off point. I hope that helps.
Regards. Jonathan
Using SPSS: Unfortunately, I know of no way to get this from the Regression dialog. But what you could do is the following, assuming we are happy with the differences between partial and part correlations:
Part Correlation significance: For each independent variable in your regression you have a zero, partial, and part correlation listed. Why not regress that independent variable (which becomes a new dependent variable) on all other independent variables, but save the residuals: Analyse -> Regression -> Linear -> Save -> Unstandardised Residuals. This will create a new variable in your SPSS datafile. That variable effectively represents the unaccounted for variance.
Now correlate that new variable with your original dependent variable. Analyse -> Correlation -> Bivariate. The significance listed in that analysis will be associated with the part correlation. You would do the above for each independent variable.
If you are interested in the same analysis but for the partial correlations you would need to regress your original dependent variable on all independent variables except the one you are testing the partial correlation significance of. Oh, and save the residuals. The newly saved variable is the dependent variable without the shared variance from the independents.
I do hope this helps. Regards. Jonathan.