What does this mean in logistic regression

noetsi

Fortran must die
#1
From a state study

The “impact” of major factors (labor market ,type of disability, type of economic/ medical assistance,
and type of DVR services received) and other client characteristics (for example, education, distance
from office) were estimated in terms of their ability to explain the likelihood of successful
rehabilitation.

Statistically, these likelihoods are derived from odd ratiosoften referred to as ‘odds’ – higher or
lower chances of an event (rehabilitation).Likelihoods are calculated by taking the odds minus 1.00.
Odds are calculated by taking the exponential value of the relevant ‘log-odd’coefficientor sum of
relevant log-odd coefficients – referred toas log-odds. They are estimated by a logistical regression

The formulas are:
Likelihood = (Odds – 1.00) Odds = Exp [Sum of Log-odds]
So: Likelihood = (Exp [Sum of Log-odds] –1.00)

For example: A post secondary degree is associated with a log-odd of 0.3223. See model. The odds
ratio is the exponential value of 0.3223:1.38. The likelihood is 0.38 (1.38 minus 1.0); it is interpreted
as a 38 percent increase in the chances of being successfully rehabilitated if a DVR client has such a
degree.

This description seems like relative risk to me, but I did not think that was calculated this way.
 

hlsmith

Not a robit
#2
Wording definitely seems weird at times but it is ORs. I think they are trying to say an OR = 1.5 is a 50% greater odds for outcome than the reference group.