Timeline for Is hierarchical regression appropriate for running a regression using multilevel dependent variable?
Current License: CC BY-SA 3.0
13 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jun 20, 2016 at 16:31 | comment | added | usεr11852 |
Yeah, It was the bit connecting from the separate regressions to the hierarchical model. Your comment made it clear. (+1). @user120062: MCMCglmm allows for multiple responses too if you are interested in MCMCing this model out.
|
|
| Jun 20, 2016 at 12:01 | history | edited | Robert Long | CC BY-SA 3.0 |
Added info about `ASReml`
|
| Jun 20, 2016 at 11:57 | comment | added | Robert Long |
@user120062 OK, but do you have students clustered within schools, class or similar ? ? You can't divide the DV into different "levels", in the sense of levels in a hierarchical/mixed/multilevel model. If you want to run such a model then you need to have a single DV, with most software that I am aware of. The only exception I know is asreml which I have just come across but have never used.
|
|
| Jun 20, 2016 at 10:41 | comment | added | user120062 | continued....My dependent variable is student score which is an aggregate score( weighted score) of the three subjects (Maths, English and Aptitude). I am not quite sure if multilevel regression is appropriate as @Robert Long suggested. To make it clear, what I am suggested is running a single model using hierarchical regression instead of running a regression for the aggregate score and then a sub-model for each subject. Of course it could not be a logistic regression as my dependent variable is a percentage not "0" and "1". | |
| Jun 20, 2016 at 10:41 | comment | added | user120062 | Thank you all for your comments. I understand how to run a a hierarchical regression by clustering the student scores by classroom, school and location etc. My concern is whether hierarchical regression is appropriate to run a regression by dividing the dependent variables into different levels. | |
| Jun 19, 2016 at 23:40 | comment | added | Robert Long | @usεr11852 maybe my answer isn't clear. Which part is the problem for you ? The last paragraph ? This would obviously only apply if there were actual hierarchies in the data such as classrooms or schools. The 2nd paragraph suggests an SEM model with multiple outcomes, while the 1st says that you can't have multiple outcomes in a hierarchical model and if you aggregate them then there won't be any clustering anyway. | |
| Jun 19, 2016 at 19:07 | comment | added | usεr11852 | Yeah, that's what I am saying too. The idea of clustering within student seems implausible unless you change the response variable to be the aggregate vector of the subject-specific scores. The way the post is written doesn't really clear that up because you mention the word model for different models. | |
| Jun 19, 2016 at 18:47 | comment | added | Robert Long | @usεr11852 glad you like it ! | |
| Jun 19, 2016 at 18:37 | comment | added | Robert Long | @usεr11852 If the student is a random effect / cluster , then what is clustered within each student ? If the 3 scores per student are aggregated into one then there would be only 1 measurement per student in the model and hence no clustering. | |
| Jun 19, 2016 at 17:52 | comment | added | usεr11852 |
One could "theoretically" reformulate the model so he aggregates the individual lesson scores into one response vector. Then he would be able to use student as a random effect but he would have to also use an indicator variable for each specific subject-specific score to model the mean response. This could probably work but you would likely bleed out some serious degrees of freedom from the model.
|
|
| Jun 19, 2016 at 17:42 | comment | added | usεr11852 | I am not sure this is advice is very accurate. A hierarchical model at student level with the response variable being the composite test score would have a different random effects level per measurement. (I like your avatar by the way, personal hero of mine that guy...) | |
| Jun 19, 2016 at 11:35 | history | edited | Robert Long | CC BY-SA 3.0 |
added 67 characters in body
|
| Jun 19, 2016 at 11:13 | history | answered | Robert Long | CC BY-SA 3.0 |