Linear Regression Assumptions

I am studying the effect of religious cleavages on the duration of a civil war. My dependent variable is the duration of a war (measured in years), my independent variable is religious cleavages, it is dichotomous (1-0) and I have created a dummy variable for it. I have then proceeded to do a linear regression (which I assume is the right test I should be using?) but when I want to check for the assumptions, I am unsure as to how to interpret the graphs.
Checking for linearity: (see attached screenshot), there is clearly no linearity, does that mean the model I am using is not the right one? What about homoscedasticity assumption?

And how do I check for normality? I did Q-Q Plots but Im not sure this is the right way: (results attached in screenshots n2 and 3)

Thank you.



Less is more. Stay pure. Stay poor.
Well the normality is in the residuals from the model. I am thinking your attachments are not for residuals?

Count based models are good for counts, but given the higher values for years, linear regression should be a reasonable substitution.


Fortran must die
If you have a non-linear relationship then you need to run a non-linear model. Unfortunately none are particularly simple.

the qq plots should be for the residuals. If you are using the qq plots for residuals that is the right way to do it.