I made a land value analysis using a random-intercept multilevel model with 3 levels (lands nested in districts further nested in cities). The intercept is allowed to vary between each district and each city.

I transform my dependent variable (USD/sqm) with log-transformation. The whole model will look like this:

log(Price)_ijk = beta0 + beta1*X1 + beta2*X2 + v_k + u_jk + e_ijk
  • beta0 = intercept
  • X1 and X2 = predictors
  • v_k = random effect from city k
  • u_jk = random effect from district j in city k
  • e_ijk = residual of land i in district j in city k

I have computed the result from v_k and u_jk respectively. Let's say, city A has district 1A and 2A, and city B has district 3B and 4B:

  • beta0 = 23.50
  • v_A = 0.77
  • u_1A = 0.08
  • u_2A = -0.10
  • v_B = 0.13
  • u_3B = 0.01
  • u_4B = -0.05

My questions would be:

  • How would I interpret these random effects, particularly random effects from city A and city B? Right now the DV is log-transformed, but I want to see how many USD/sqm is the difference of effect coming from city A and city B to the land value (not the log-transformed value).
  • Can I interpret the result without considering the effects coming from the districts within their cities? Currently, I am only interested in seeing the effect coming from each city.

Thank you for your attention!


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.