"Good controls", also known as “confounders” or “deconfounders”, are variables which are theorized to be unaffected by the treatment and which are intended to eliminate omitted-variable bias.
[1] "Bad controls", on the other hand, are variables that could be affected by the treatment, might contribute to collider bias, and lead to erroneous results.
Scientific experiments test the relationship of an IV (or independent variable: that element that is manipulated by the experimenter) to the DV (or dependent variable: that element affected by the manipulation of the IV).
[5][4] This situation may be resolved by first identifying the confounding variable and then redesigning the experiment taking that information into consideration.
[4][5] Take, for example, the well known combined gas law, which is stated mathematically as: where: In an experimental verification of parts of the combined gas law, (P * V = T), where Pressure, Temperature, and Volume are all variables, to test the resultant changes to any of these variables requires at least one to be kept constant.