EViews.com

Hi,

i have a doubt about the estimation of the predicted values ​​of the TOBIT model.

I estimated a tobit equation with this coefficients:

Coef. Std Err. t P> | t |
uc_conjunt .0000131 3.91e-06 0001 -3.36
Rain .0000944 .0000608 0122 1.55
salary .0003148 .0000509 6.18 0.000
C .745279 .132751 5.61 0.000
Scale .353847 .0176239


Analyzing the EViews Help i noticed that there are two ways to estimate the predicted values: Index - Expected latent variable and Expected dependent variable.

What is the difference between them?

Using the first line of my data, i obtained the results below. How i obtain the corrected result manually?

data:
Rain salary uc_conjunt
1685 2985.773332 20742.71001

Results from EViews:
expected 1.579478185664588
latent 1.571544924599041

Algotithm:
tobit: EFFICIENCY = C + + SALARY + RAIN UC_CONJUNT, left truncated = 1
std = Scale
prev = X' * Beta
lambda = cdf(prev / std) / pdf (prev / std)
Latent = pdf (prev / std) * (prev + (lambda * std))

The results coincide with the latent variable (1.571544924599041). I would like to estimate Expected dependent variable (1.579478185664588).

How to manually calculate this?

Thank you

Fernando Cunha

The manual has the generic formula for the expected dependent variable (Equations 26.29 - 26.31).You just have to substitute the probabilities for the classifications using your CDF and the truncated expected value of the latent. The latter is the most involved. You can look at the representations view to see the expressions.