
Explain the statement
that an experiment is a model. Note and explain the similarities and
differences between an experimental model and a theoretical model of the type
that you have studied exclusively in your courses.
 Exp model – relationship among parameters w/ physical
 simulations
Similarity – both are relationships btwn parameters
 Differences – Experimental data vs. mathematical
 relationships to generate model

Explain what it means
that an experiment should produce results that correlate to performance in the
real world. Explain how we can attempt to assure that this is the case?
 Experiments should produce results that closely resemble
 what happens in the real world; good experiments give results that are similar
 to field data
 To ensure similarity btwn experiment data and real world,
 try to use same conditions, same materials, same correlations

The other side of
designing a good experiment is to evaluate & critique an experiment. State
2 aspects of exp design, conduct, or reporting that could be used to attack exp
validity and explain how they might be used to question results.
Exp design:
Too much uncertainty
Not controlling parameters well
Confounding variables
Reporting:
Conclusions stretch beyond scope of work
Bad data reduction

Explain difference
and importance: Error & Uncertainty
 Error  % difference from known true value; provides definite
 conclusion of “right/wrong”
 Uncertainty – amount of variation w/ respect to accuracy;
 confidence of conclusions in exp

Explain difference
and importance: Findings & Conclusions
Findings – summary of data trends and observations
 Conclusions – Answer objectives; statements of understanding
 that can be applied to predict results of other experiments

Explain difference
and importance: Objectives & Tasks
 Objectives – explicit statement of what is to be done,
 determined, answered, or developed; gives explicit purpose for doing the
 experiment
 Tasks – Things that need to be done to accomplish the
 objectives

Explain difference
and importance: Goal & Objectives
Goals – more broad, long term things to be accomplished
 Objectives – explicit statement of what is to be done,
 determined, answered, or developed; gives explicit purpose for doing the
 experiment

Explain difference
and importance: Correlation & Causation
 Correlation – Some trend exists between two variables, but
 may be nonquantifiable; an experiment might be discarded if no correlations
 exist
 Causation – Quantifiable proof that one variable has a
 direct effect on the behavior of another variable; may legitimize an experiment
 by demonstrating data trends

Explain difference
and importance: Rectification of results & Normalization of results
 Rectification – using mathematical manipulation to present
 data as a straight line
 Normalization – comparison of data to some baseline to
 establish a measure of relative difference

Good or Bad Obj: “The
objective of this report is to compare toaster ovens a,b,c,d”
Bad; “compare” is too vague.
 “The objective of this report is to compare the cooking
 times of toasters a,b,c,d”

Good or Bad Obj: “The
objective was to measure the time AA batteries last at different loads”
Bad; “measure” is a task.
 “The objective of the exp is to determine the rate of electrical
 discharge of AA batteries at different loads”

Statisticians usually
define uncertainty using the resulting data for some type of regression
analysis. This is:
 “Not useful during planning” & “useful after experiment
 is completed even if it was a single value”

Extraneous variables
are:
 “continuous variables which are secondary independent
 variables”
 &
 “variables external to an experiment which may affect the
 outcome”

The effect of
extraneous variables may be suppressed by:
 “controlling the experiment so that extraneous variables
 have little effect”
 &
 “randomizing data taking so that the effect of
 extraneous variables is spread over the test envelope”

The spacing of data
points within the test envelope:
 “should be done to equalize the uncertainty throughout the
 test envelope”

For a 3 factor
experiment with 4 levels of factor 1, 3 levels of factor 2, and 2 levels of
factor 3, what number of tests will be required to conduct a full factorial
design?
24 = 4*3*2

Consider a 3 factor
experiment with 4 levels of factor 1, 3 levels of factor 2, and 2 levels of
factor 3. How many terms will there be in a linear multiple regression equation
for the result including all effects?
8

For 2nd
order multi regression equation for 2 independent variables, how many terms
will be in equation considering all effects?
2

For a 3 factor
experiment, how many significance tests will be required to do ANOVA?
7=3!+1

Number of significant
figures to be used for reporting numerical results should:
“Convey accuracy of instruments”
& “be based upon measurement with worst uncertainty”
& “reflect # of sig figs used for calc”
& “be based upon uncertainty of value reported”

Uncertainty analysis
should be used:
“when reporting results”
& “during planning of exp”
& “during the execution of an experiment”

To graphically rectify
data exhibiting exponential behavior, ___ coord. Should be used
Semilog

When plotting data
for pressure drop for fluid flow in duct, where uncertainty is equal throught
test envelope; the data should be spaced at:
Equal increments of flow rate (x variable)

When designing an
experiment one should strive for uncertainty which is:
Adequate, considering what results will be used for

When designing an
experiment the amount of data that is planned to be collected should be
 Just
 enough to plot a curve

An irreversible
experiment always requires a ____ plan
Sequential

An efficient design
of an experiment results in
“elimination of extraneous variables”
& “minimum uncertainty”
& “adequate useful data”

A random experiment
design should be used:
To suppress the effect of extraneous variables

A classical
experiment plan
Involves holding one variable constant at a time

Adequate uncertainty
refers to uncertainty that is:
Sufficiently low so that the objective can be satisfied

An irreversible
experiment is one for which:
Measurements cannot be repeated because of a transition

ANOVA can be used to
determine
 The confidence level of the significance of the effect of a
 factor on the response variable

Randomized block
experiment designs can be used
To suppress the effect of discrete extraneous variables

Balance equations can
be used
 “To check on the reasonableness of data”
 &
 “establish
 criteria for rejecting data”

Questions concerning
your results should be
 Anticipated and considered during the design of the
 experiment

An experiment is
confounded when
 The effect of one independent variable on the dependent
 variable is indistinguishable from the effect of another

