Designing Controlled Experiments

NOTE SHEET 2 – Designing
Controlled Experiments
Designing Controlled Experiments




Scientists can solve problems and test theories by
designing and performing controlled experiments.
In a controlled experiment, the scientist chooses
something to test (an independent variable) to see if
it causes a change in something else that they will
measure (the dependent variable).
The experiment is controlled because the scientist makes
sure that all experimental variables are kept constant
and only changes the independent variable.
By doing this, the scientist makes sure that the
dependent variable can only change if the independent
variable causes it to do so.
Designing Controlled Experiments

For example, a scientist could see if caffeine (IV)
affects growth (DV) in mice
A controlled experiment has 4 basic parts:
1. THE PROBLEM – What are you trying to find
out?

Independent Variable (IV) – What is tested
to see if it has an effect on something else

Dependent Variable (DV) – What is affected
or changed because of the IV – The DV in a
controlled experiment is what you measure as
your data

Problem Statement – Question that relates
the IV to the DV

How will the
IV
affect the
DV ?
Hypothesis


A hypothesis is a prediction based on prior
research about what will happen during the
experiment
For the problem “How will caffeine affect growth
in mice?” there are a few possible hypotheses:



The caffeine will have a negative effect on the
growth of the mice
The caffeine will have a positive effect on the
growth of the mice
The caffeine will have no effect on the growth
of the mice
A controlled experiment has 4 basic parts:
2. THE EXPERIMENTAL DESIGN (or
PROCEDURE) – What are you going to do to
investigate the problem?

The procedure is the steps you are going to
follow to do the experiment

A good procedure is:



clearly written in numbered steps
as precise and exact as possible (include
measurements, units, etc.)
easily repeated by somebody else
Why does the procedure have to be
as precise and exact as possible?




Consider the following cake recipe:
Ingredients: sugar, butter, eggs, vanilla extract,
flour, baking powder, milk
Directions: Heat oven and grease pan. Mix
sugar and butter, add eggs and stir in vanilla.
Combine flour and baking powder and add to
sugar mixture. Add milk and stir. Pour batter
into pan and bake.
What is wrong with this recipe?
A Better Recipe
Directions
1.
Ingredients
 1 cup white sugar
 ½ cup butter
 2 eggs
 2 teaspoons vanilla
extract
 1 ½ cups all purpose
flour
 1 ¾ teaspoons
baking powder
 ½ cup milk
2.
3.
4.
5.
Preheat oven to 350 degrees F
(175 degrees C). Grease and
flour a 9x9 inch pan or line a
muffin pan with paper liners.
In a medium bowl, cream
together the sugar and butter.
Beat in the eggs, one at a time,
then stir in the vanilla.
Combine flour and baking
powder, add to the creamed
mixture and mix well.
Finally stir in the milk until batter
is smooth. Pour or spoon batter
into prepared pan.
Bake for 30 to 40 minutes in the
preheated oven. For cupcakes,
bake 20 to 25 minutes. Cake is
done when it springs back to the
touch.
Mmmm…Cake
Experimental Design Continued



Constant – Variables in the experiment
that remain the same each trial
Control – A test group without the IV
used for comparison to see if the IV really
has any affect on the DV
Repeated Trials – Doing tests more than
once to get more accurate data
Does shampoo
work?




If you ALWAYS wash your
hair with shampoo, how
do you know if it actually
works?
If you try washing your
hair with just water, and
it looks and feels the
same, does the shampoo
do anything?
Would you know this
without trying just water?
In this example, using
water only would be
considered a control –
this way you would know
if the shampoo works
better or worse than
nothing at all
Why repeat
trials?




If you flip a coin one time
and get heads, is this a
correct conclusion:
Every time you flip a coin,
you get heads
If you flip the coin a few
more times, eventually
you will get a tails
If you get a result once, it
is meaningless, but if the
experiment is repeated
and the results are
always the same, then
that result is reliable
A controlled experiment has 4 basic parts:
3. THE DATA – A measure of the IV’s effect on
the DV

Data is collected and presented in a Data
Table AND a Graph
Data Table and Graph


In a data table, the IV goes in the left
column and the DV goes in the right
column
In a graph, the IV is on the X-axis
(bottom) and the DV is on the Y-axis (left
side)
A controlled experiment has 4 basic parts:
4. THE CONCLUSION – An answer to the
problem

A conclusion uses data to answer the
question asked by the problem
statement and discusses the
experiment’s validity

A good conclusion is always supported
by data from the graph and data table
Why do I need to back up the
conclusion with data?


Consider the following
statement by itself:
Michael Jordan is the
greatest basketball
player in the history
of the NBA
The Same
Statement
With Data







Michael Jordan is the greatest
basketball player in the
history of the NBA
NBA 50th Anniversary AllTime Team (1996)
Member of six NBA
championship teams (199193, 1996-98)
Five-time NBA Most Valuable
Player (1988, 1991-92, 1996,
1998)
Six-time NBA Finals Most
Valuable Player (1991-93,
1996-98)
Ten-time All-NBA First Team
(1987-93, 1996-98)
All-NBA Second team (1985)
Michael Jordan is
the greatest…





Nine-time NBA AllDefensive First Team
(1987-93, 1996-98)
NBA Defensive Player of
the Year (1988)
NBA Rookie of the Year
(1985)
NBA All-Rookie Team
(1985)
Two-time IBM Award
winner, for all-around
contribution to team's
success (1985, 1989)




Three-time NBA All-Star
Game MVP (1988,
1996, 1998)
Participated in 11 NBA
All-Star Games (1985,
1987-1993, 1996-98),
starting 10 times, and
missed another due to
injury
Holds the NBA All-Star
Game career record for
highest scoring average
(21.3 ppg)
Recorded only tripledouble in All-Star Game
history, with 14 points,
11 rebounds and 11
assists, in the 1997 NBA
All-Star Game in
Cleveland
Michael Jordan is
the greatest…



Won the Nestle Crunch
Slam Dunk in 1987 and
1988, also participating in
1985
Passed Kareem AbdulJabbar to become the
NBA's all-time leading
playoff scorer (5,762
points) during the 1998
Eastern Conference Finals
Passed Dennis Johnson
into fourth place on the
NBA's all-time playoff
assists list (1,006), in an
88-83 victory over the
Indiana Pacers in Game 7
of the Eastern Conference
Finals in 1997-98



Scored his 29,000th
career point, posting a
game-high 41 points, 6
rebounds and 4 assists,
in a 107-93 win over
the Minnesota
Timberwolves on 4/3
Broke Kareem AbdulJabbar's NBA record by
scoring in double-digits
for the 788th
consecutive game,
scoring a game-high 33
points, against the
Minnesota
Timberwolves on
12/30/97
Scored a career-high 69
points in a 117-113
overtime win at
Cleveland on 3/28/90
The Conclusion Continued - Validity


Validity – How reliable are the results of
an experiment?
Results are more reliable, or valid, when:



the experiment is repeated more than once
(repeated trials)
The experiment is carefully designed and
performed
The experiment keeps all variables constant
(except for the IV)
The results of an experiment are
not valid if…




All the variables in the experiment are not
carefully controlled
The procedure does not include
measurements
The experiment is done only one time
If mistakes are made during the
experiment that might affect the results
Experimental Scenario


A science teacher heard a rumor that students
listening to classical music perform better on
tests. He decides to test this theory when he
gives his next test to his class, which contains
20 students. Design a controlled experiment
to test this theory.
Directions First, state the problem and
identify the variables. Second, write a
numbered procedure that shows how the
experiment will be performed.
Experimental Scenario




Problem Statement
How will listening to classical music affect
the student’s performance on tests?
IV is listening to classical music
DV is student’s performance on tests