Cross-sectional data - MLM

9 important questions on Cross-sectional data - MLM

How do you compare the variances of 2 models with eachother when you want to use variance to explain the differences in outcome between clustervariables?

Model 1 = intercept only model + random intercept (1 cluster variable)
model 2 = add variable to fixed part + random intercept (1 cluster variable)

(variance model 1 - variance model 2) / variance model 1 = % of the difference in outcome variable between clusters explained by the variable added.

How do you use ICC as indicator of explained variance? 2 steps

1. Make an intercept-only model. So only with the outcome variable and cluster variable (so no determinant)
2. Determine the ICC

What is the formula for ICC with linear regression?

This.
Var (_cons) is between variance
Var (_cons) + var (residual) = total variance.
so between variance/total variance
  • Higher grades + faster learning
  • Never study anything twice
  • 100% sure, 100% understanding
Discover Study Smart

What is the interpretation of using variance as explanation? (model 1 random intercept only model, model 2 random intercept + determinant)

X% of the difference in outcome between [clustervariable] is explained by this variable added to the model in fixed part.

Why dont you have a residual variance with logistic regression?

Because it says something about probabilitys, and you dont have residual variance when you say something about probabilities?

How do you use variance as an indicator?

1. Make a model 1 with only random intercept (cluster on 1 level)
2. Write down the variance of random intercept cluster 1
3. Make a model 2 with cluster on 2 levels
4. Write down the changed variance of random intercept cluster 1
5. Compare the changed variance

What is the formula for variance as indicator?

(variance [clustervariable1] in model 1 - variance [clustervariable1] in model 2) / variance [clustervariable1] in model 1

What is the interpretation when variance is used as an indicator? So when for example variance in neighborhood goes down by adding another level of clustering(country)?

X% of clustering in neighborhood is caused by clustering in country.
so then you can say (when you for example have 70%), most of the explanation is not because of neighborhood but because of country.

Why is the residual variance higher when you have 2 levels of clustering in stead of 1?

With 2 levels, the residual variance is a combination of errors at country level(cluster 2) and errors at neighborhood level (cluster 1)

The question on the page originate from the summary of the following study material:

  • A unique study and practice tool
  • Never study anything twice again
  • Get the grades you hope for
  • 100% sure, 100% understanding
Remember faster, study better. Scientifically proven.
Trustpilot Logo