A between-Subject psychology experiment. The Independent Variable is distant tie (i.e., some one not close to me, a stranger) vs. close tie (i.e., someone I feel close to). The close tie condition is further broken down into "brother" and "son," which caused the trouble. The DV is related with choice. My hypothesis is when we choose products for a close tie, we value different things compared to when we choose products for a distant tie.
My ultimate goal is to test Tie---Mediator---DV mediation. So seems I should pool "brother" and "son" into one "close tie" condition, but pooling test (i.e., chow test) is significant meaning that they can not be pooled together using raw data values. This is because "brother" and "son" differ significantly regarding the DV. So, seems using residuals as the new DV is the only option, right? If so, could you check if my plan is correct? Thank you soooo much!
Step 1. Include only close tie cases(i.e., son cases + brother cases). Use the son/brother independent variable (possible value: "son" coded as 1, and "brother" coded as 0). DV is the original DV, sth related with product choice. Save "standarized residuals."
Step 2. Include ALL cases. Compute the standarized value of the original DV. Retain the standardized values for "distant tie" cases and discard those for "close tie" cases.
Step 3. Combine "standarized residuals" generated at step 1. and "standarized value" of distant cases at step 2. into one new variable. Use this new variable as the new DV in analyses that include all cases.
I feel my 3-step procedure inappropriately shrinks the difference between the distant tie condition and the other two conditions...
Or, there is a procedure for this in some software package instead of me computing all the residuals & standardized values? I use SPSS, though I guess the type of software doesnt matter in this case. Thank you!
My ultimate goal is to test Tie---Mediator---DV mediation. So seems I should pool "brother" and "son" into one "close tie" condition, but pooling test (i.e., chow test) is significant meaning that they can not be pooled together using raw data values. This is because "brother" and "son" differ significantly regarding the DV. So, seems using residuals as the new DV is the only option, right? If so, could you check if my plan is correct? Thank you soooo much!
Step 1. Include only close tie cases(i.e., son cases + brother cases). Use the son/brother independent variable (possible value: "son" coded as 1, and "brother" coded as 0). DV is the original DV, sth related with product choice. Save "standarized residuals."
Step 2. Include ALL cases. Compute the standarized value of the original DV. Retain the standardized values for "distant tie" cases and discard those for "close tie" cases.
Step 3. Combine "standarized residuals" generated at step 1. and "standarized value" of distant cases at step 2. into one new variable. Use this new variable as the new DV in analyses that include all cases.
I feel my 3-step procedure inappropriately shrinks the difference between the distant tie condition and the other two conditions...
Or, there is a procedure for this in some software package instead of me computing all the residuals & standardized values? I use SPSS, though I guess the type of software doesnt matter in this case. Thank you!
Last edited: