Hello,
I'm trying to conduct a post-hoc test following ANOVA, comparing mean mental health score (mh_score) by type of abuse (abuse4). The ANOVA shows significant difference in means across groups:
anova mh_score abuse4 [weight=screen_wt]
(analytic weights assumed)
(sum of wgt is 1,847)
Number of obs = 1,903 R-squared = 0.1074
Root MSE = 8.50713 Adj R-squared = 0.1060
Source | Partial SS df MS F Prob>F
-----------+----------------------------------------------------
Model | 16529.945 3 5509.9817 76.13 0.0000
|
abuse4 | 16529.945 3 5509.9817 76.13 0.0000
|
Residual | 137433.07 1,899 72.371284
-----------+----------------------------------------------------
Total | 153963.01 1,902 80.947957
Notice I use survey weight in this analysis. However, when I use the -tukeyhsd- or -tukeykramer- or -tkcomp- commands, they are not compatible with weights. And they use the unweighted means to make the comparison. How do I use the weighted means in this comparison?
Thanks.
I'm trying to conduct a post-hoc test following ANOVA, comparing mean mental health score (mh_score) by type of abuse (abuse4). The ANOVA shows significant difference in means across groups:
anova mh_score abuse4 [weight=screen_wt]
(analytic weights assumed)
(sum of wgt is 1,847)
Number of obs = 1,903 R-squared = 0.1074
Root MSE = 8.50713 Adj R-squared = 0.1060
Source | Partial SS df MS F Prob>F
-----------+----------------------------------------------------
Model | 16529.945 3 5509.9817 76.13 0.0000
|
abuse4 | 16529.945 3 5509.9817 76.13 0.0000
|
Residual | 137433.07 1,899 72.371284
-----------+----------------------------------------------------
Total | 153963.01 1,902 80.947957
Notice I use survey weight in this analysis. However, when I use the -tukeyhsd- or -tukeykramer- or -tkcomp- commands, they are not compatible with weights. And they use the unweighted means to make the comparison. How do I use the weighted means in this comparison?
Thanks.
