Key DoE Terms

Key DoE Terms

Your Cheat Sheet for talking DoE

Factors, factor levels, response variable, … there are some terms that you will encounter over and over in DoE. The table below contains the most important ones together with a short definition and an example. Once you familiarize yourself with those terms, you can read the following chapters without getting lost, and we can continue to the next chapter to learn how we design trustworthy experiments using randomization, replication, and blocking.

Essential DoE Terms

TermDefinitionExample
FactorThe variables you want to investigate
in your experiment.
The effect of temperature on the yield of a
chemical reaction. Temperature is a factor.
You can include more like pressure,
concentration, …, as many as you want.
Disturbance
Variable
The variables that you do not want to
investigate but still might influence
your experiment.
The impact of the batch of the raw material.
Factor RangeThe specified limits or boundaries within
which the factors are to be evaluated.
The material decomposes above 250 °C, and
the reaction needs at least 120 °C to start.
Therefore, a meaningful temperature range
is set from 120 °C to 250 °C.
Factor LevelThe actual values you are going to test
within the factor range.
Within the 120 °C to 250 °C range, you can
test many temperatures, but you usually
choose the extremes (120 °C and 250 °C)
and sometimes one in the middle (185 °C).
Those are the factor levels.
Response
Variable
The outcome that is measured.The yield is your response variable.
Again, you can have more than one.
ReplicationRepeating the same experiment multiple
times to account for experimental error
and uncontrollable disturbance variables.
You determine how the yield of a chemical
reaction varies with temperature. Then, you
repeat the test at 150 °C three times. This
will give you an idea of the variability
in your measurements.
RandomizationConducting experiments in a random
order to avoid bias.
Instead of testing temperatures in a
sequential order (e.g., 100 °C, 150 °C,
200 °C), you can test them in random order
(e.g., 150 °C, 200 °C, 100 °C).
BlockingBlocking in experimental design involves
organizing experiments in groups (blocks)
to minimize errors caused by known
disturbance variables that cannot or
are not controlled.
You have two different batches of a
material and while you not really
interested in the difference of these
two batches they might influence
your experiment.
VarianceMeasure of the spread of data points
around the mean.
Variance tells you how consistent your
measurements are. If the variance is small,
the three yields at 200 °C are close
together—like 95%, 95.5%, and 96%.
If the variance is large, the values are
more spread out—like 85%, 92%, and 98%.
ResidualsDifference between observed and
predicted values in a model.
Residuals tell you how far off your
model’s prediction is from the actual
result. If your model predicts a yield
of 95% at 200 °C but you measure 96%,
the residual is +1%. Small residuals mean
your model fits the data well—large ones
mean it’s missing something.
ANOVAAnalysis of Variance; tests whether
differences between group means are
statistically significant.
If you measure yields at 120 °C and 250 °C;
ANOVA checks if the changes in yield that
you observe are real or just happened
by chance.
SignificantMeans the result is unlikely to be due
to chance and likely reflects a real effect.
If the yield at 250 °C is significantly
higher than at 120 °C, it suggests
temperature truly affects the reaction
and the variation was not random.
P-valueOutput of ANOVA. Indicates how likely
it is that your observations are by chance.
If you measure yields at 120 °C and 250 °C,
and ANOVA outputs a p-value of 0.04,
the likelihood that the observed difference
in yield was due to chance is only 4%.

Up next:

<< Principles of DoE: Randomization, Replication, Blocking >>

<< Main Effects & Interaction Effects Explained >>