Not sure to fully understand your question. In general, the intercept is an estimate of the population mean for individuals with qualitative explanatory variables at their reference value, and quantitative explanatory variables at 0. The consequence is that you must make sure that the ref value or 0 values make sense. For instance, if you have the body mass as a predictor, you should centre it on some meaningful value before doing the regression.
Thanks for the answer...I apply MRA to estimate homes' values so the dependent variable is the value of residential properties and the independent variables are area, age and floor..I try to interpret the model and I can't understand what really means in my case the constant??
What is the nature of the dependent variable ? If it is not quantitative, you're not in the field of regular MRA. If it is, the meaning of the constant is the value of residential properties for zero values of area, age and floor. As I explained above, if the independent variables are quantitative, they should be centred before running the analysis, e.g. by substracting their sample mean (or some other meaningful value). When independent variables are centred on their mean, the intercept is an estimate of the population mean value of residential properties for an average area, average age and average floor.
Multiple linear regression model (full model for quantitative response variable Y as ‘value of residential properties and ‘n’ predictors x1, x2, ..., xn as area, age, floor, etc.) can be expressed as:
Y = α + β1x1 + β2x2 + .... + βnxn
Here, the constant term ‘α’ represent the average value of response variable in absence of all predictors. Mathematically, that’s correct. However, a zero setting for all predictors in a model is sometimes an impossible/nonsensical combination.
As in your case, it does not make sense to state the average value of residential property with area, age, floors, etc. all set to be zero [we can have zero age – recently constructed property, but non logical to interpret house with no floors, and no area]
For detailed discussion on interpreting constant term in MRA see: http://blog.minitab.com/blog/adventures-in-statistics/regression-analysis-how-to-interpret-the-constant-y-intercept
I recommend that you check to make sure that the relationships are in fact linear. The variables you mention often have something more like a lognormal distribution, which means that you should do multiple linear regression on the log-transformed variables. However, it is possible to calculate a regression with a zero intercept ("regression through the origin") if you are sure that Y must be zero when all the Xs are zero.