The coefficients in the model represent the impact of each variable on the model. They are not statistical measures. Moreover, they are arbitrary, in the sense that they are directly sensitive to scale. For example, in a model that included voltage as a predictor variable, if the model's coefficient were, say, 2.7 in volts, it would be 2700 in millivolts. For this reason, when as often the variables are in incommensurate units, it can be helpful to normalize each variable (e.g. (x - mean(x))/std(x)).