logistic regression with non boolean variables I have trouble wrapping my head around the concept of logistic regression and its interpretation for non-Boolean variables. 
An example from our course material goes as follows:

Calculating a generalized linear model in R for the formula
  liked(boolean)~ highCalories(boolean), the coefficients look like this
Coefficients:  

             Estimate  Std. Error   z value    Pr(>|z|)   
(Intercept)  -0.47367    0.04369    -10.841    < 2e-16  
 Calorie s   -0.71497    0.17398     -4.109    3.97e-05 





With the intercept being -0.47, the chance of liked being 1 with
  highCalories being 0 is logit(-0.47) -> 38.4%
If highCalories is 1, the chance of liked being also 1  is
  logit(-0.47367-0.71497) -> 23.4%

So in the case of the examined variable being Boolean it's clear to me. Let's look at an example with non-Boolean variables:

Coefficients:  

             Estimate  Std. Error   z value    Pr(>|z|)   
(Intercept) -3.806e+00  2.801e-01   -13.589    <2e-16   
day         -1.934e-03  1.862e-04   -10.387    <2e-16

this is for MailIsRefound(boolean)~dayofstudy(numeric)
Because dayofstudy is not Boolean, the logit for the intercept doesn't seem that meaningful, but for day we get logit(-3.806e+00 -1.934e-03) -> 2.1%
How is this value to be interpreted? 
 A: It simply means that for each unit change in the predictor day (presumably the unit is 'days') the log odds ratio of the response changes by a step of $-1.934\times 10^{-3}$, i.e. the odds ratio of the response changes by a factor of $\exp(-1.934\times 10^{-3})= - 0.99880619$.  So for each change in the predictor of a week, the log odds ratio of the response should change by a step of $7\times-1.934\times 10^{-3}=-0.01358$, i.e. the odds ratio of the response should change by a factor of $\exp(-0.01358)=-0.9865118$.  Of course it's worth checking that this presumed linear relationship between the predictor & the log odds ratio of the response seems reasonable; just as with ordinary least squares regression, mutatis mutandis.
A: The exponential of the parameters in a logistic regression can be interpreted in terms of Odds Ratio (p/1-p)/(p'/1-p'). You can view this link:
http://www.unm.edu/~schrader/biostat/bio2/Spr06/lec11.pdf
As you can see, exp(b)=OR for a binary covariate, and exp(b)^Lambda=OR of augmenting the numerical covariate in Lambda units.
