Hello everyone,
I have a technique question: "For a multilevel model, how are the coefficients of indicators from different levels calculated?" and need your help. Hope to receive your answers!
I understand the basic rules of the multilevel model and under which situations we should use it. However, how the coefficients are calculated is still unclear to me.
For the lower-level indicators, the Stata manual shows that the model estimates a linear model for each group, and gets an overall coefficient by giving weights to the coefficients of each separated model. Thus I guess for groups in which a variable does not show any variation, they hardly contribute to the calculation of the coefficient of this variable. Is my understanding correct?
Regarding the coefficients of higher-level indicators, I did not find explanations. Does the calculation regard the group mean of the dependent variable as a virtual dependent variable and calculate the influence of group attributes like a simple linear regression does? In this case, the observation number is the group number.
Here is my dataset, which has a high-level categorical indicator, showing clear differences in descriptive analysis and in a simple linear regression (I know the data does not fulfil the requirement of IID, and it is only for comparison). However, this indicator is insignificant in the multilevel model. I am not clear about the reason for this difference. I attached the tables.
If my previous understanding is wrong, can anyone provide me with some answers or related documents?
Many thanks!
Hong Yan
Descriptive analysis
The linear regression model
The reference of the weather sensitivity group is the “Sensitive group”
Similar to descriptive analysis, the less-sensitive group and especially the less-rain-sensitive group cycle slower than the sensitive group.
The three-level model with only the weather-sensitivity group
The weather sensitivity group is not found to influence cycling speed
I have a technique question: "For a multilevel model, how are the coefficients of indicators from different levels calculated?" and need your help. Hope to receive your answers!
I understand the basic rules of the multilevel model and under which situations we should use it. However, how the coefficients are calculated is still unclear to me.
For the lower-level indicators, the Stata manual shows that the model estimates a linear model for each group, and gets an overall coefficient by giving weights to the coefficients of each separated model. Thus I guess for groups in which a variable does not show any variation, they hardly contribute to the calculation of the coefficient of this variable. Is my understanding correct?
Regarding the coefficients of higher-level indicators, I did not find explanations. Does the calculation regard the group mean of the dependent variable as a virtual dependent variable and calculate the influence of group attributes like a simple linear regression does? In this case, the observation number is the group number.
Here is my dataset, which has a high-level categorical indicator, showing clear differences in descriptive analysis and in a simple linear regression (I know the data does not fulfil the requirement of IID, and it is only for comparison). However, this indicator is insignificant in the multilevel model. I am not clear about the reason for this difference. I attached the tables.
If my previous understanding is wrong, can anyone provide me with some answers or related documents?
Many thanks!
Hong Yan
Descriptive analysis
The linear regression model
The reference of the weather sensitivity group is the “Sensitive group”
Similar to descriptive analysis, the less-sensitive group and especially the less-rain-sensitive group cycle slower than the sensitive group.
The three-level model with only the weather-sensitivity group
The weather sensitivity group is not found to influence cycling speed
