Experimental Hypotheses

1
Experimental Hypotheses
- specific hypotheses about what would happen in
a particular situation are generated by good
theories
- these hypotheses must be testable
- requirement for operational definitions
- issue of construct validity
2
- if we use self-report as our operational
definition of depression, we need to make sure
that self-report is a valid indicator of the
underlying construct (i.e., asking people how
they feel is a good way to measure actual levels
of depression)
- we usually do this by comparing results
obtained using our measure with results obtained
using other accepted operational definitions of
the construct
- if people who are classified as depressed by
our measure are classified as normal by other
accepted measures (or vice versa), there is
reason to question the construct validity of our
operational definition
3
correlation with other measures is an indicator
of construct validity in most cases
4
Experimental Hypotheses
Hypotheses and Theory
established theories usually generate hypotheses
using deductive reasoning
established theory all mammals are
warm-blooded
hypothesis 1 cats are warm-blooded
hypothesis 2 dogs are warm-blooded
hypothesis 3 goats are warm-blooded
5
Experimental Hypotheses
Hypotheses and Theory
new theories are usually generated using
inductive reasoning
fact 1 cats have tails
fact 2 cats are warm-blooded
fact 3 dogs have tails
fact 4 dogs are warm-blooded
new theory all warm-blooded organisms have
tails
6
Experimental Hypotheses
Hypotheses and Theory
new theory all warm-blooded organisms have
tails
how would you now use deductive reasoning to test
this new theory?
e.g., hypothesis all monkeys have tails
when deduced hypotheses show that a theory cant
be true, we have to examine the inductive
reasoning that generated it
7
Experimental Hypotheses
Testing Hypotheses
Strong Inference
- when you design an experiment to test a
hypothesis, you need to consider other possible
explanations/hypotheses
- ideally, the experiment should be designed so
that these other hypotheses are explicitly tested
(and hopefully ruled out)
8
Experimental Hypotheses
Null and Alternative Hypotheses
You have invented Drug X, which you believe will
prevent heart attacks
- the experiment you design will actually test
the hypothesis that Drug X will have no effect on
the incidence of heart attacks
- this is the null hypothesis (the manipulation
of the IV will have no effect on the DV)
9
Experimental Hypotheses
Null and Alternative Hypotheses
You have invented Drug X, which you believe will
prevent heart attacks
- if your data do not support the null
hypothesis, this hypothesis is rejected in favor
of the alternative hypothesis
alternative hypothesis
Drug X does have an effect on the incidence of
heart attack
10
Experimental Hypotheses
Null and Alternative Hypotheses
Directional Predictions
- we may want to be more specific about just what
effects we think our drug will have
(specifically, it will reduce heart attacks - -
the alternative hypothesis)
- therefore, the null hypothesis that we test
should be Drug X will have no effect on, or
increase, the incidence of heart attack
11
Experimental Hypotheses
Null and Alternative Hypotheses
Directional Predictions
- we may expect an effect in a particular
direction (e.g., DV will increase)
- we may expect an effect, but be unsure of the
direction (e.g., DV may increase or decrease)
- how we frame our hypotheses has statistical
implications down the road
12
Experimental Hypotheses
13
Experimental Hypotheses
Where Ideas Come From
Four Stage Model
1. preparation
2. incubation
3. illumination
4. verification
14
Experimental Hypotheses
Where Information Comes From
1. Books (can be found in libraries)
2. Academic/Professional Journals (can also be
found in libraries)
3. Internet (much caution required)
15
Experimental Hypotheses
Searching the Literature
there are a number of large databases that can be
searched for journal articles and/or books on a
particular topic
- in psychology, the 2 most frequently used are
Psychinfo and Medline (both available online
through WLU library)
- many journals are also available online (older
issues may not be)
16
Scales of Measurement
In order to conduct any experimental science, we
will eventually have to measure something
Four basic scales of measurement
1. Nominal Scales
2. Ordinal Scales
3. Interval Scales
4. Ratio Scales
17
Scales of Measurement
1. Nominal Scales
- a nominal scale measures something in terms of
quality, rather than quantity
- also referred to as categorical scales
- no implication that any one category is more
or less than any other, they are just different

• e.g., male/female

• e.g., jazz/country/techno/death metal

18
Scales of Measurement
2. Ordinal Scales
- ordinal scales are used when data are
quantifiable on some continuum
- can be used to indicate that one value is
greater than/less than another value, but not by
how much
- e.g., finishing 1st, 2nd, 3rd, etc... in a race
- e.g., house numbers
19
Scales of Measurement
3. Interval Scales
- data are quantifiable on some continuum, and
intervals between consecutive values are equal
- the difference between 2 and 5 is the same size
as the difference between 626 and 629
- e.g., temperature in C or F
- in these scales, 0 does not mean there is no
temperature (zero is arbitrary)
20
Scales of Measurement
4. Ratio Scales
- data are quantifiable on some continuum,
intervals between consecutive values are equal,
and zero really means zero (actually represents
the quantity of zero)
- e.g., height, weight
- e.g., temperature in K
21
Next Lecture
- more on reliability and validity
- Chapter 7 (control)
- Assignment 1 available