Dear Forum
I am running a simple random intercept multilevel model with Observations nested in Individuals who are nested in demand sets (4), who are in turn nested in a condition (4) that is nested in the treatment (5 control +4) . The treatments are coded as dummy variables. I also have a bunch of control variables. I ran the model using the command:
mixed ocf_brt_adjustment_score total_ferrw_4p i.graph_usage i.performance_metrics_1 i.total_fanchart_knowledge i.comp_att_check bias_c0n40 brt_c0n40 fan_c0n40 deco_c0n40 base_c0n40 bucket_dum1 bucket_dum2 bucket_dum3 bucket_dum4 bucket_dum5 bucket_dum6 bucket_dum7 bucket_dum8 bucket_dum9 if condition =="C0N40" || treatment_code : || con_flag : || demand: || un_name :
I am including a sample output below. The control treatment base_c0n40 is understandably omitted because of collinearity.
I am having trouble interpreting the coefficients. If I take the brt_c0n40 highlighted in red the value of the co-efficient is 0.14 . Do I read that as compared to the base_c0n40 the brt_c0n40 delivers an increase of 0.14 in the dependent variable?
If so, then the margins command shows something different. The margins command shows that the value of the dependent variable is 0.29 for the brt_c0n40 treatment and 0.46 for the base_c0n40. Meaning the dependent variable is actually lower for the brt_c0n40 treatment. Shouldnt the co-eeficient be negative?
I appreciate any feedback.
Thank you for your help
I am running a simple random intercept multilevel model with Observations nested in Individuals who are nested in demand sets (4), who are in turn nested in a condition (4) that is nested in the treatment (5 control +4) . The treatments are coded as dummy variables. I also have a bunch of control variables. I ran the model using the command:
mixed ocf_brt_adjustment_score total_ferrw_4p i.graph_usage i.performance_metrics_1 i.total_fanchart_knowledge i.comp_att_check bias_c0n40 brt_c0n40 fan_c0n40 deco_c0n40 base_c0n40 bucket_dum1 bucket_dum2 bucket_dum3 bucket_dum4 bucket_dum5 bucket_dum6 bucket_dum7 bucket_dum8 bucket_dum9 if condition =="C0N40" || treatment_code : || con_flag : || demand: || un_name :
I am including a sample output below. The control treatment base_c0n40 is understandably omitted because of collinearity.
I am having trouble interpreting the coefficients. If I take the brt_c0n40 highlighted in red the value of the co-efficient is 0.14 . Do I read that as compared to the base_c0n40 the brt_c0n40 delivers an increase of 0.14 in the dependent variable?
If so, then the margins command shows something different. The margins command shows that the value of the dependent variable is 0.29 for the brt_c0n40 treatment and 0.46 for the base_c0n40. Meaning the dependent variable is actually lower for the brt_c0n40 treatment. Shouldnt the co-eeficient be negative?
I appreciate any feedback.
Thank you for your help
