How we load our data sets with theories and why we do so purposefully

•Theory-laden observations need not be treated as a nuisance. They can be created purposefully for their epistemic benefits.•Some scientific inferences cannot be understood unless we understand how and why our observations are theory-laden.•The theories loaded onto our observations need not play a role in the justification of our scientific models.

In this paper, I compare theory-laden perceptions with imputed data sets. The similarities between the two allow me to show how the phenomenon of theory-ladenness can manifest itself in statistical analyses. More importantly, elucidating the differences between them will allow me to broaden the focus of the existing literature on theory-ladenness and to introduce some much-needed nuances. The topic of statistical imputation has received no attention in philosophy of science. Yet, imputed data sets are very similar to theory-laden perceptions, and they are now an integral part of many scientific inferences. Unlike the existence of theory-laden perceptions, that of imputed data sets cannot be challenged or reduced to a manageable source of error. In fact, imputed data sets are created purposefully in order to improve the quality of our inferences. They do not undermine the possibility of scientific knowledge; on the contrary, they are epistemically desirable.