In MLM analyses, school types (public or private) at level 2 were highly correlated with school predictors (at the same level - L2), when using it in the model it suppressed all L2 predictors. It has no effect on any covariates at L1 (i.e., student covariates). Therefore, I decided to use the school predictors (this is the research questions ref.) without adding to the model the school types and all worked fine. How can I still use the school types? Knowing that the L2 predictors are constructs in the data and converted to composites by using EFA. While the school type is a sampling stratum!
Multicollinearity is a phenomenon caused by a high correlation between two or more predictor variables in a multiple regression model. In this situation the coefficient estimates of the multiple regression may change erratically in response to small changes in the model or the data.
In the first place, you have to evaluate if that is what is happening in your model. For example, you may evaluate if large changes in the regression coefficients occur when a predictor variable is added or deleted. In addition you can calculate the variance inflation factor (VIF). Values higher than 5 strongly suggest a colinearity problem.
Several strategies have been described to deal with multicollinearity. However, the most practical (and which I prefer) are:
1. Preselect or maintain in the model only the variables that are really relevant and that provide differential information. In this sense, if you have two predictors that are highly correlated with each other (for example, with a correlation coefficient of 0.6 or greater) choose only one that has a greater conceptual relevance and / or stronger association with the outcome.
2. If possible, obtain more data. More data can produce more precise parameter estimates and, therefore, more stable models.
Although I do not know your study, I suppose that the type of school is highly correlated with school predictors. Therefore, it is likely that the school type variable does not contribute additional information and therefore, the decision to exclude it from the model may be valid.
Multicollinearity is a phenomenon caused by a high correlation between two or more predictor variables in a multiple regression model. In this situation the coefficient estimates of the multiple regression may change erratically in response to small changes in the model or the data.
In the first place, you have to evaluate if that is what is happening in your model. For example, you may evaluate if large changes in the regression coefficients occur when a predictor variable is added or deleted. In addition you can calculate the variance inflation factor (VIF). Values higher than 5 strongly suggest a colinearity problem.
Several strategies have been described to deal with multicollinearity. However, the most practical (and which I prefer) are:
1. Preselect or maintain in the model only the variables that are really relevant and that provide differential information. In this sense, if you have two predictors that are highly correlated with each other (for example, with a correlation coefficient of 0.6 or greater) choose only one that has a greater conceptual relevance and / or stronger association with the outcome.
2. If possible, obtain more data. More data can produce more precise parameter estimates and, therefore, more stable models.
Although I do not know your study, I suppose that the type of school is highly correlated with school predictors. Therefore, it is likely that the school type variable does not contribute additional information and therefore, the decision to exclude it from the model may be valid.
Dear YashinYou may opt for factor analysis. This will reduce the dimension of the data and makes your issue simple.
Regards
Thanks Fredi, option 1 was what I did ... I excluded the school type from the model for 2 reasons: the predictors in my study are the school level factors (e.g., resources, leadership, ... etc.), and the second is what you mentioned. I wanted someone to support or critique my approach, and this is what you did. Thanks again!
Multicollinearity is the occurrence of high inter-correlations among independent variables in a multiple regression model.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS