Hi, I know that the model that fits better with my data is a Count Data Model (CDM) but I would like to know how can I justify the use of this model instead of OLS or other lineal model. (Basically I know that my data fits with CDM because my dep. variable is a number of events and this kind of data is modeled in that way in the literature).
I would like to know how can I justify that.
Thank you in advance!
You can reason for example the following way: Since OLS assumes normally distributed residuals, OLS model cannot rule out e.g. negative values. Therefore, the fitted or predicted value from OLS could be negative, whereas you know that in your dependent variable is never negative.
Hi
The following papers may be of help to you.
Thank you and have a wonderful time
In the case of Count Outcomes the estimates of linear Regression can be inconsistent or inefficient.
I agree on most answers although I would say it may depend on the values taken by the dependent variable. If the distribution of the dependent variables has a large mean but a small scatter, then using OLS will be acceptable. I am more familiar with time series where there are numerous count data models and many do not allow regressors.
Can multilevel Poisson regression be used to study TSH disease?
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS