I'm trying to predict a continuous variable using multiple linear regression model including various categorical and continuous independent variables. In the model I have also included the "location" variable (which includes the categories, i.e., name of the location/intersection from where the data were gathered).

I'm wondering whether in a fixed effect model the “location” could be used as an independent variable to see location wise influence, or should I fit a random intercept linear model to account for the influence of "location".

To understand it, I have fitted a random intercept linear model, to account for the random effect of "location". After fitting the model, I have calculated the variance explained by the random effect parameter (location), which revealed that it only explained 12.83% of the variance.

My queries are:
1. Should I proceed with random intercept linear model or due to the low variance explained by the model I should stick with fixed effect linear model?
2. If I stick with fixed effect MLR, then is it ok to add “location” as independent variable. What I observed that during fitting the fixed effect MLR model, removing location variable cause significant change in other variables; thus I found it as important variable in the model.

Any suggestion regarding this query will be appreciated.

Thanks