If you want to conduct a scientific experiment, you need to
do the following four steps:
1.
Determine the broad question(s) that you want to
answer
->Formulate research questions
or hypotheses
2.
Determine the different treatment conditions
needed
->Translate the research
hypotheses into a set of treatment conditions
3.
Select the experimental design by which to test
the participants in the treatment conditions
->How are the different
treatments administered? Different groups of subjects or participants?
4.
Record the response measure that you are
interested in.
Experimental Variables
Independent variable:
=>Experimental or treatment variable, something that
varies between the treatment conditions, TEST the IV
=>Quantitative IV-variation in terms of amount (amount of
drug, loudness of noise)
=>Qualitative IV-variations in type or strategy (type of
meal)
=>Classification IV-variations that are intrinsic to the
subjects of the experiment (age, sex, IQ, species, word type…)
Dependent variable:
=>Measured or response variable, MEASURE the DV
=>Something that can “capture” the hypothesized
differences, somehow depend on the IV
=>DV could depend on many things than the IV (important
to control these other influences)
Nuisance variables:
=>DVs left uncontrolled which could exert a systematic
influence (affects DV) on different treatment conditions (time of day, nature
of participants)
=>When a NV is not controlled, it becomes a confounding
variable
Experimental Designs
1.
Between-groups design: completely randomized
design
>participants are randomly assigned to serve in one of the treatment
conditions
>works for most nuisance variables
>differences in behaviour observed among the treatment conditions are
based on differences among different groups of participants/independent
individuals
2.
Within-subjects design: randomized block design
>uses blocks of participants
who are matched closely on some relevant characteristics
>treat a participant as a “block”,
where the individual serves in all treatment conditions of an IV
>also known as repeated–measures
design/within-subjects design
Statistical vs. Research Hypotheses
**Research Hypothesis=general experimental statement about
the assumed nature of the world (broad statement linking IV and DV)
**Research Hypothesis=set of precise hypotheses about the parameters
of the different treatment conditions
-Null (H0) and
Alternative(H1/HA) Hypotheses: mutually exclusive or
incompatible
-H0
says that there are no treatment effects in the population, i.e. changes in IV
DO NOT affect the DV, the values of a particular parameter is same in different
treatment populations
-When the parameters obtained
are too different from those specified by H0, H0 is
rejected in favour of the H1.
-H1 usually states
that the values of the parameter in the different treatment populations are not
all equal, i.e. there is an effect of IV on DV.
*H0 is not rejected
if he parameter estimates are close to those specified by it
You cannot prove anything! You always collect evidence for
or against. You are always testing the H0. You are always trying to
find evidence against the H0 (so you may reject it in favour of H1).
It is impractical to obtain populations means-->we have
sample means.
(Sample is representative of the overall population, usually
as large as feasible.)
α<Alpha>: the criteria for rejecting H0
Errors in hypothesis testing:
- · Type 1 error
o
Decision=reject the H0 & accept
the H1
o
Reality=H0 should not have been
rejected.
o
Cause: α-level too lenient
- · Type 2 error
o
Decision=fail to reject H0 &
accept it
o
Reality=H1 is true & H0
should have been rejected
o
Cause: α-level too stringent, confounding
variable
>>If it is more important to discover new facts, we
may be more willing to acceot more type 1 error; if it is more important to not
clog up the literature with false facts, then we might be more willing to accept
more type 2 errors.
没有评论:
发表评论