Time taken to receive funding: A count variable? I am trying to predict time taken by a startup to receive funding based on specific characteristics of entrepreneur. Regardless of a bunch of transformations and using linear or mixed effects model, the residuals do not nicely fit normality. I understand that there are generalized linear models for such purposes. But can time taken to receive funding (in days, hours or minutes) be conceptually treated as a "count" variable so that poison or negative binomial models can be fit?    
 A: Time to an event (outcome variable) is usually analyzed using survival analysis methods - for example, Cox proportional hazards regression models if the interest is in relating this time to a bunch of predictor variables (in this case, characteristics of entrepreneurs). For your setting, the event consists of receiving funding. 
The reason for this is that not all startups will receive funding during the period of study - those who will NOT receive funding will have their time to event censored (i.e., partially observed), whereas those that will receive funding will have their time to event uncensored (i.e., fully observed).
For startups with a censored time to event, their time to event will occur sometime beyond the end of the study, but we just won't known exactly when. For these startups, the censored time to event will be set to the time to the end of study (measured from when we first started monitoring whether firms receive funding).  
Time to event data therefore consists of two pieces of information for each startup:


*

*Their time to event;

*Whether or not that time to event was censored or uncensored.

