Dear all,
I am trying to estimate a model with a three-level hierarchical structure using STATA 16’s mixed command.
Level 3: Regions
Level 2: States
Level 1: Observations (repeated over the years within states)
My model is as follows:
I introduce both random intercepts and random slopes for the independent variable FDI at the state and region level. This means that each state and region in the data set have their own intercepts, as well as the effect of FDI on the dependent variable Y is allowed to vary at the state and region level. The need for the addition of the random parameters has also been confirmed by the respective LR tests, intra-class correlations, etc.
Estimation results suggest that covariation between the variance of the random intercept and slope parameters at the region level is statistically insignificant. Put differently, random parameters do not move together and using ‘covariance(unstructured)’ option after mixed command does not have an added value for the specification. However, when I plot the values of the intercepts and slopes at the region level, they seem to be related (please see the plot below). I find myself puzzled and confused a little, and do not know how to interpret this difference.
Will you please help me? In fact, the figure shows random intercepts and slopes are related, but there is not much variation in the data. Is this the correct way of interpretation?
Many thanks in advance
I am trying to estimate a model with a three-level hierarchical structure using STATA 16’s mixed command.
Level 3: Regions
Level 2: States
Level 1: Observations (repeated over the years within states)
My model is as follows:
Code:
mixed Y A set of independent variables || region: FDI || state: FDI
Estimation results suggest that covariation between the variance of the random intercept and slope parameters at the region level is statistically insignificant. Put differently, random parameters do not move together and using ‘covariance(unstructured)’ option after mixed command does not have an added value for the specification. However, when I plot the values of the intercepts and slopes at the region level, they seem to be related (please see the plot below). I find myself puzzled and confused a little, and do not know how to interpret this difference.
Will you please help me? In fact, the figure shows random intercepts and slopes are related, but there is not much variation in the data. Is this the correct way of interpretation?
Many thanks in advance
