I like math, i like economics; but i don't like math in my economics
Hard science models allows for an estimated uncertainty based on the limitations of the measurements. Measuring the width of a board with a centimeter ruler yields a estimated uncertainty of about plus or minus 1mm. All is well with our measurement because we can estimate the percentage of uncertainty. But, measuring some type of social variable too claims an estimated uncertainty. The problem is not with the model as much as it is with the estimated uncertainty. We know what is uncertain in the How are we to know what is uncertain in the centimeter example because we know what is certain. We know the length of a centimeter simply because we know that a centimeter is 1/100 of a meter and we know that a meter is defined a the distance light travels in a vacuum during the time interval of 1/299,792,458 of a second, and we know that the second is defined as the frequency of radiation emitted by cesium atoms when the pass between two particular states (more specifically as the time required for 9,192,631,770 periods of radiation). Yet, in a social science we survey groups, establish central tendency information and establish estimated uncertainty as if we knew the initial certainty. How do we know if the regression contains a rich array of covariates? We don't. We know little.