**HELP ME! I am doing Negative Binomial Regression Model and these are the Results obtained**

[Guest]

I’m presently engaged in researching my thesis, focusing on the connection between poor mental well-being and economic progress, specifically examining its impact through the lens of absenteeism. Absenteeism, which indicates the total number of days the respondent was absent from work due to health-related issues in the last 12 months, is my dependent variable.

This is the regression that I built:

ABSENT = β0 + β1Depressed + β2FEMALE + β3AGE + β4Pre-Teriatary + β5Teriatary + β6MARRIED + β7PN0 + β8HHNBPERS13CS17 + β9HINCOME + β10FT_PTCS19 + ԑ

AW2 is the number of absenteeism days in past year.
β0 is the intercept.
Depressed is a poor mental health indicator.
Female is the gender.
AGE is the age group that the respondent falls into.
Pre-tertiary and Tertiary is the level of education.
MARRIED is the marital status.
PN0 is the physical pain suffered by the individual.
HHNBPERS13CS17 is the number of children aged 13 years or less residing in the household.
HINCOME is the household’s total net income per month.
FT_PTCS19 is a dummy variable 1 - Full time; 0 - otherwise.
ԑ is the error term.

My data is cross-sectional and case-based and due to a significant proportion of zero entries (i.e., zero absent days) I have an overdispersion. I am using STATA 13.

I am new to this and this is my first time using Negative Binomial Regression Model. Is this a good model? Can you provide me with your feedback and suggestions? Also, I am not sure how to interpret the results correctly. Below you can find the model. Thank you so much for your help in advance.

[Carlo Lazzaro]

Guest:
1) it's true that -nbreg- allows parameterizing the overdispersion detected in -poisson-;
2) at theit face value,your models give back very similar results: when it comes to choose on of them, most depends on the predictors in the right-hand side of your regression equation (which, in turn, depends on the data ganerating process). See the literature of your research field and bring this issue(s) up to your supervisor.

[Guest]

Carlo Lazzaro thank you for your message. Almost all of my independent variables are insignificant at P<0.05 hence I used P<0.10 in order to have some significant variables. Does it make sense to use P<0.10 because it is rarely to find that in a research paper? Also, should I give importance to the significance of the variables when choosing a model? Thank you

[Carlo Lazzaro]

Guest:
1) judging a model by the statistical significance of its coefficients is not that scientific. That said, you may find papers where (rarely, though) the arbitrary cut-off is set at 0.10. However, this value is as arbitrary as .05: it only doubles the chance to wrongly reject the null when it is correct;
2) much more substantively, you should double-check that your model is correctly specified (that is, all the parameters included in the data generating process are plug in the right-hand side of your regression equation).

[Nick Cox]

Cross-posted at https://stats.stackexchange.com/ques...-are-the-resul

Please note our policy on cross-posting, which is that you should tell us about it.