7.7 KiB
| title | chunk | source | category | tags | date_saved | instance |
|---|---|---|---|---|---|---|
| Scientific control | 3/4 | https://en.wikipedia.org/wiki/Scientific_control | reference | science, encyclopedia | 2026-05-05T03:45:31.111708+00:00 | kb-cron |
Y
~
⊥
A
|
U
,
X
{\displaystyle {\tilde {Y}}\perp A|\;U,X}
: There is no causal effect of
A
{\displaystyle A}
on
Y
~
{\displaystyle {\tilde {Y}}}
given
X
{\displaystyle X}
and
U
{\displaystyle U}
. The NCO is independent of the treatment given
X
{\displaystyle X}
and
U
{\displaystyle U}
. U-Comparability:
Y
~
⧸
⊥
U
|
X
{\displaystyle {\tilde {Y}}\not {\perp }U|\;X}
The unmeasured confounders
U
{\displaystyle U}
of the association between
A
{\displaystyle A}
and
Y
{\displaystyle Y}
are the same for the association between
A
{\displaystyle A}
and
Y
~
{\displaystyle {\tilde {Y}}}
. Given assumption 1 - 4, a non-null association between
A
{\displaystyle A}
and
Y
~
{\displaystyle {\tilde {Y}}}
, can be explained by
U
{\displaystyle U}
, and not by another mechanism. A possible violation of Latent Exchangeability will be when only the people that are influenced by a medicine will take it, even if both
X
{\displaystyle X}
and
U
{\displaystyle U}
are the same. For example, we would expect that given age and medical history (
X
{\displaystyle X}
), general health awareness (
U
{\displaystyle U}
), the intake of
A
{\displaystyle A}
influenza vaccine will be independent of potential influenza related deaths
Y
~
A
=
a
{\displaystyle {\tilde {Y}}^{A=a}}
. Otherwise, the Latent Exchangeability assumption is violated, and no identification can be made. A violation of Irrelevancy occurs when there is a causal effect of
A
{\displaystyle A}
on
Y
~
{\displaystyle {\tilde {Y}}}
. For example, we would expect that given
X
{\displaystyle X}
and
U
{\displaystyle U}
, the influenza vaccine does not influence all-cause mortality. If, however, during the influenza vaccine medical visit, the physician also performs a general physical test, recommends good health habits, and prescribes vitamins and essential drugs. In this case, there is likely a causal effect of
A
{\displaystyle A}
on
Y
~
{\displaystyle {\tilde {Y}}}
(conditional on
X
{\displaystyle X}
and
U
{\displaystyle U}
). Therefore,
Y
~
{\displaystyle {\tilde {Y}}}
cannot be used as NCO, as the test might fail even if the causal design is valid. U-Comparability is violated when
Y
~
⊥
U
{\displaystyle {\tilde {Y}}{\perp }U}
, and therefore the lack of association between
A
{\displaystyle A}
and
Y
~
{\displaystyle {\tilde {Y}}}
does not provide us any evidence for the invalidity of
A
{\displaystyle A}
. This violation would occur when we choose a poor NCO, that is not or very weakly correlated with the unmeasured confounders.
== Positive control ==
Positive controls are often used to assess test validity. For example, to assess a new test's ability to detect a disease (its sensitivity), then we can compare it against a different test that is already known to work. The well-established test is a positive control since we already know that the answer to the question (whether the test works) is yes. Similarly, in an enzyme assay to measure the amount of an enzyme in a set of extracts, a positive control would be an assay containing a known quantity of the purified enzyme (while a negative control would contain no enzyme). The positive control should give a large amount of enzyme activity, while the negative control should give very low to no activity. If the positive control does not produce the expected result, there may be something wrong with the experimental procedure, and the experiment is repeated. For difficult or complicated experiments, the result from the positive control can also help in comparison to previous experimental results. For example, if the well-established disease test was determined to have the same effect as found by previous experimenters, this indicates that the experiment is being performed in the same way that the previous experimenters did. When possible, multiple positive controls may be used—if there is more than one disease test that is known to be effective, more than one might be tested. Multiple positive controls also allow finer comparisons of the results (calibration, or standardization) if the expected results from the positive controls have different sizes. For example, in the enzyme assay discussed above, a standard curve may be produced by making many different samples with different quantities of the enzyme.
== Randomization ==
In randomization, the groups that receive different experimental treatments are determined randomly. While this does not ensure that there are no differences between the groups, it ensures that the differences are distributed equally, thus correcting for systematic errors. For example, in experiments where crop yield is affected (e.g. soil fertility), the experiment can be controlled by assigning the treatments to randomly selected plots of land. This mitigates the effect of variations in soil composition on the yield.
== Blind experiments ==