How does sense-data pose a challenge to an empiricist account of scientific practice? And so how might one remedy this?
1) The first means to overcome errors, which are due to a wrong interpretation of sense data, is to repeat the observation under different conditions, to change the experiment, and to verify the result by several groups of experimentators. The latter serves to eliminate the subjective impact of the observer.
By these means certain illusions can be detected, e.g., the illusion of the stick in water, which is observed as broken.
2) But there remain wrong interpretations of sense data which cannot be resolved this way, e.g., the observation that the sun rises in east and sets in west.
Here one has to change the ambient theory. History shows that it can be a long way to overcome this kind of illusion and to detect, that it is not the sun which moves but the earth. These errors can only be detected and corrected in the context of a whole theory, not by focussing solely on the observation in question.