Hi,
I am investigating a variable for its effect as a mediator with SPSS and Process but I have problems interpreting the output.
First I examined the following relationship: IV1 -> M -> DV
My IV is not significant on M. But M is significant on my DV. So neither the total effect nor the direct effect are significant. I would conclude that M predicts DV but that IV1 does not predict neither M nor DV. But when I look at the indirect effect and the Bootstrap confidence intervals, and calculating the difference between upper and lower bound, it is 0,102, so it is different from zero implying that the indirect effect is significant. This result is contradicting to the other insignificant effects. How to interpret the output correctly? Is there a mediation?
Run MATRIX procedure:
************* PROCESS Procedure for SPSS Release 2.16.3 ******************
Written by Andrew F. Hayes, Ph.D. www.afhayes.com
Documentation available in Hayes (2013). www.guilford.com/p/hayes3
**************************************************************************
Model = 4
Y = DV
X = IV1
M = QM
Sample size
376
**************************************************************************
Outcome: QM
Model Summary
R R-sq MSE F df1 df2 p
,069 ,005 1,295 1,802 1,000 374,000 ,180
Model
coeff se t p LLCI ULCI
constant 3,217 ,084 38,352 ,000 3,052 3,382
IV1 ,158 ,117 1,343 ,180 -,073 ,388
**************************************************************************
Outcome: DV
Model Summary
R R-sq MSE F df1 df2 p
,197 ,039 1,244 7,521 2,000 373,000 ,001
Model
coeff se t p LLCI ULCI
constant 4,160 ,183 22,783 ,000 3,801 4,519
QM ,176 ,051 3,466 ,001 ,076 ,275
IV1 ,173 ,115 1,496 ,135 -,054 ,399
************************** TOTAL EFFECT MODEL ****************************
Outcome: DV
Model Summary
R R-sq MSE F df1 df2 p
,088 ,008 1,280 2,942 1,000 374,000 ,087
Model
coeff se t p LLCI ULCI
constant 4,726 ,083 56,647 ,000 4,562 4,890
IV1 ,200 ,117 1,715 ,087 -,029 ,430
***************** TOTAL, DIRECT, AND INDIRECT EFFECTS ********************
Total effect of X on Y
Effect SE t p LLCI ULCI
,200 ,117 1,715 ,087 -,029 ,430
Direct effect of X on Y
Effect SE t p LLCI ULCI
,173 ,115 1,496 ,135 -,054 ,399
Indirect effect of X on Y
Effect Boot SE BootLLCI BootULCI
QM ,028 ,024 -,008 ,094
Partially standardized indirect effect of X on Y
Effect Boot SE BootLLCI BootULCI
QM ,024 ,021 -,007 ,081
Completely standardized indirect effect of X on Y
Effect Boot SE BootLLCI BootULCI
QM ,012 ,011 -,003 ,040
Ratio of indirect to total effect of X on Y
Effect Boot SE BootLLCI BootULCI
QM ,138 1,855 -,114 1,351
Ratio of indirect to direct effect of X on Y
Effect Boot SE BootLLCI BootULCI
QM ,160 18,823 -,283 4,547
R-squared mediation effect size (R-sq_med)
Effect Boot SE BootLLCI BootULCI
QM ,002 ,003 ,000 ,012
Normal theory tests for indirect effect
Effect se Z p
,028 ,023 1,209 ,227
******************** ANALYSIS NOTES AND WARNINGS *************************
Number of bootstrap samples for bias corrected bootstrap confidence intervals:
1000
Level of confidence for all confidence intervals in output:
95,00
NOTE: Kappa-squared is disabled from output as of version 2.16.
------ END MATRIX -----
Second, I checked another variable for a possible mediation (IV2 -> M -> DV) and got a different output.
This time the influence from IV2 on M is significant as well as the influence of M on DV. IV2 is not significant on M. I would conclude from that, that I have a full mediation. But the total effect model is not significant. So maybe I don't have a mediation? Does anyone know that? (The interaction effect with Bootstrap again is significant.)
Run MATRIX procedure:
************* PROCESS Procedure for SPSS Release 2.16.3 ******************
Written by Andrew F. Hayes, Ph.D. www.afhayes.com
Documentation available in Hayes (2013). www.guilford.com/p/hayes3
**************************************************************************
Model = 4
Y = DV
X = IV2
M = QM
Sample size
376
**************************************************************************
Outcome: QM
Model Summary
R R-sq MSE F df1 df2 p
,371 ,138 1,122 59,801 1,000 374,000 ,000
Model
coeff se t p LLCI ULCI
constant 2,855 ,079 36,061 ,000 2,699 3,010
IV2 ,846 ,109 7,733 ,000 ,631 1,061
**************************************************************************
Outcome: DV
Model Summary
R R-sq MSE F df1 df2 p
,191 ,037 1,247 7,094 2,000 373,000 ,001
Model
coeff se t p LLCI ULCI
constant 4,229 ,177 23,951 ,000 3,882 4,576
QM ,205 ,055 3,760 ,000 ,098 ,312
IV2 -,147 ,124 -1,188 ,236 -,392 ,097
************************** TOTAL EFFECT MODEL ****************************
Outcome: DV
Model Summary
R R-sq MSE F df1 df2 p
,011 ,000 1,290 ,049 1,000 374,000 ,826
Model
coeff se t p LLCI ULCI
constant 4,814 ,085 56,702 ,000 4,647 4,981
IV2 ,026 ,117 ,220 ,826 -,205 ,257
***************** TOTAL, DIRECT, AND INDIRECT EFFECTS ********************
Total effect of X on Y
Effect SE t p LLCI ULCI
,026 ,117 ,220 ,826 -,205 ,257
Direct effect of X on Y
Effect SE t p LLCI ULCI
-,147 ,124 -1,188 ,236 -,392 ,097
Indirect effect of X on Y
Effect Boot SE BootLLCI BootULCI
QM ,173 ,062 ,075 ,316
Partially standardized indirect effect of X on Y
Effect Boot SE BootLLCI BootULCI
QM ,153 ,054 ,064 ,274
Completely standardized indirect effect of X on Y
Effect Boot SE BootLLCI BootULCI
QM ,076 ,027 ,032 ,137
Ratio of indirect to total effect of X on Y
Effect Boot SE BootLLCI BootULCI
QM 6,704 2,36E+012 2,806 7,47E+013
Ratio of indirect to direct effect of X on Y
Effect Boot SE BootLLCI BootULCI
QM -1,175 23,110 -123,620 1,332
R-squared mediation effect size (R-sq_med)
Effect Boot SE BootLLCI BootULCI
QM -,004 ,008 -,021 ,009
Normal theory tests for indirect effect
Effect se Z p
,173 ,052 3,359 ,001
******************** ANALYSIS NOTES AND WARNINGS *************************
Number of bootstrap samples for bias corrected bootstrap confidence intervals:
1000
Level of confidence for all confidence intervals in output:
95,00
NOTE: Kappa-squared is disabled from output as of version 2.16.
------ END MATRIX -----
Thanks a lot for your help!
Tweet |