# Interpretation of Expression in ordinal regression

Suppose we are looking at an ordinal variable with $4$ categories. So there are three threshold coefficients $b_1, \dots b_3$ and one probit slope $b_4$. What is the interpretation of the following expressions: $$\frac{b_{1}}{b_{4}^{2}}, \frac{b_{2}}{b_{4}^{2}}, \frac{b_{3}}{b_{4}^2}, \frac{1}{b_{4}}$$

assuming a multivariate normal distribution?

-
In what sense are you assuming a multivariate response distribution? Typically we think there is a univariate normal response distribution that has been categorized by replacing the value of the latent variable with an ordinal category based on where the value is relative to the thresholds. –  gung Jan 4 '13 at 20:16
–  whuber Jan 4 '13 at 21:36
Why are you looking at these expressions? The threshold divided by the square of the slope? Why would that be interesting? –  Peter Flom Jan 4 '13 at 23:23