BIOT 6214 NAME Ketaki Pathak
Assignment #8
5 points
1. (1 point)
a) What are some considerations/assumptions that you want to make before designing an
experiment? Provide two assumptions.
Answer: There are two important considerations/assumptions that need to be made before designing
an experiment:
1. You need to "Define your Objective(s)" and determine What question are you trying to answer
(Response) before starting the experiment design. This helps establish clear direction and purpose for
your experimental work.
2. You must include all variables that may have an impact on the outcome since missing any variable
may lead to wrong conclusion.
b) If you have data for an independent variable and a dependent variable but base on your initial
data the independent variable doesn’t seem to have an effect although you are expecting that
the opposite conclusion. What would you do to see if your own conclusion maybe statistically
acceptable?
Answer: If the independent variable doesn't seem to influence the dependent variable despite expecting
the opposite, you should consider expanding your data collection range. When data is collected within a
limited range, the effect might be "buried" in the noise, making it difficult to detect any significant
impact. By expanding the range wide enough (while keeping it meaningful and within reason), the effect
can become more apparent compared to the noise.
2. (1 point) An experiment with 4 factors and 3 levels each factor. What is the total number of
possible factors-levels combinations for a full factorial design?
Answer: For a full factorial design that includes all possible factor-level combinations, when we have k
factors and each factor has m levels, the total number of possible factor-level combinations is calculated
as n = m^k. In this case, with 4 factors (k=4) and 3 levels (m=3) for each factor, the total number of
possible factor-level combinations would be:
n = 3 × 3 × 3 × 3 = 34 = 81
3. (1 point) What is the difference between confounding (aliasing) variable(s) and lurking variable?
Answer: Confounding (aliasing) occurs when source(s) of the effect cannot be identified because two or
more factors change together. For example, in a study of back pain treatment, when Ibuprofen and Dry
Needling are always administered together or not at all, it becomes impossible to determine which
treatment is causing the pain relief effect, or if both are responsible. In contrast, lurking variables are
variables that we may not be aware of, but they often have impact on outcome (error). To minimize the
impact of lurking variables, the design process includes randomizing the order of runs in the experiment.
This distinction is important because confounding is a known limitation in the experimental design
where we can identify the variables but cannot separate their effects, while lurking variables are
unknown factors that can introduce error into our results without our awareness.
4. (1 point) What is the biggest flaw of one-factor-at-a-time design of experiment?
Answer: One-Factor-At-a-Time (OFAT) design is fundamentally flawed because it ignores all interactions
between factors. This was considered a backward thinking from the last century based on the belief that
"At some point of the study you have to do OFAT to know where the effect comes from." The OFAT
approach, where only one factor is changed at a time while keeping all other factors unchanged, is very
inefficient for data collection. It does not collect sufficient information for adequate data interpretation
and may not even save your number of runs when the number of factors is large.
5. (1 point) If you have a study with 4 continuous factors, 2 level each and 1 response variables.
Suppose you are interested in finding out which factor(s) and 2-factor interactions have
significant effect(s) on your result. How would use DOE in JMP and perform fractional factorial
design since performing a full factorial design is not approved for your experiments? Show a
screenshots of your design list as well as your final table DOE.
