Transcript Research review

Research Process

Parts of the research study




Aim: purpose of the study
Target population: group whose behaviour is
investigated
Procedure: step-by-step process used by
researcher to carry out the study
Findings: states how the researcher
interpreted the data collected
Research

Measurement and Sampling
How to describe what we observe is the
main question.


The use of ratio variables (measurements
based on a continuous scale with an obvious
zero point)
Researchers want measurements to be
reliable and valid
Measurements

Sampling and population concerns

Researchers must be concerned that a sample is
representative of the population




Researchers will use random sampling and crosssectional sampling for the best results (the difficulty lies
in making the random sample representative)
Other problems with samples are: self-selected samples;
convenience samples
Cross-sectional sampling is a deliberate selection of
subjects to make the sample representative
Opportunity Sampling? Okay to use? What about
for you as an IB student?
Research design
Hypothesis
Research method
Observations
Variables
All of the aforementioned are part of the
research design.
Pitfalls in Experimental Research

Internal validity: maintaining consistent
conditions in the experimental situation,
and ensure no unwanted factors creep in

Confounds: a situation where two variables
change simultaneously making it impossible
to determine their relative influence

Ways to avoid confounds
1.
2.
Hold constant factors which are not of direct interest
Use multiple independent variables when the variables
involved are of direct interest
Pitfalls in Experimental Research

Bias: systematic error
Subject bias
 Experimenter bias
 These situations can be avoided by using singleblind and double-blind designs (deception)

Observation to Interpretation

Statistics
Concerned with the description and
interpretation of scientific data

Used to describe and summarize results


Descriptive stats
Assist in understanding what the results mean

Inferential stats
Descriptive Stats

Frequency distribution


Rearranging the scores in order of size and
then see how many people got each score
Can provide a clearer picture of what data
looks like
Descriptive Stats

Central Tendencies




Mode: most frequently occurring
Median: the middle of the frequency data
Mean: sum of all scores divided by the number of
scores
Normal and skewed distributions


Bell shaped curve is a normal distribution; highest point
occurs in the middle of the distribution
Lopsided distribution is skewed, the mean is unlikely to be
representative of the majority; in most cases, the researcher
will prefer the median as a way of describing the typical
result
Measures of Variability

Variability tells us how the scores are
distributed around the center



One indicator is the range from lowest to
highest
The next indicator is to find the deviation
scores
Squaring the deviation scores will give the
variance, however, this gives inflated data

Finding the square root of the mean of the
variance will give the standard deviation
Standard Deviation

The standard deviation will provide a
measure of variability which reflects the
position of every score within the group,
expressed in the same units as the original
scores.

The larger the standard deviation, the greater
the variability of scores
Normal Distributions

What do they tell researchers?
Knowing something is distributed “normally” tells
researchers specific things, that most scores are near
the mean and that very few scores are away from
the center (very little standard deviation)
(Look at page 449 in your handout, figure A.6)
 Knowing these properties of normal distributions
becomes very useful in making predictions about
scores and in our ability to interpret the results of
research

Correlations


Any relationships between variables are
correlational and do not directly identify
causal factors
Correlation or Causation
Two types of correlational patterns
1.
2.
Positive: occurs when increases in one variable are
associated with increases in the other variable
Negative: occurs when increases in one variable
occur as the value of the other variable decreases
(turn to page 453 in the handout, figure A.8)
Correlations

Correlational patterns are measured using
a statistical measure called a correlation
coefficient


This is a number between 0.0 and +1.0 for
positive correlations; -1.0 and 0.0 for negative
correlations
As the value moves from 0 to the maximum
the degree of the relationship between the
variables becomes stronger
Inferential Stats
Inference: a logical conclusion based on
what I know
“In using inferential stats, we try to
generalize from our sample to the
population.” (Glassman and Hadad, 2004)
Sampling and Variability

Sampling and Variability




The recognition that not all samples will be alike and
that any sample may differ from the population is the
result of sampling variability
Therefore, nothing is set in stone. All research has
sampling variability, as a consequence, more research
should be done with other samples to prove theory
There are many questions that arise when doing
research, especially when a researcher is trying
to interpret data.
“Inferential statistics are concerned with
providing guidelines” for evaluation
Inference with Normal Distributions


The simplest situation for inference
Looking at a single score in relation to a
set of data
Significance

Results which are interpreted as based on
a real effect are referred to as
significant.


The statistical tests for evaluating the chance
versus the real effects are called
significance tests
“The conclusion one draws, expressed as the
probability that the outcome is due to chance,
is called the significance level of the
results.” (Glassman and Hadad, 2004)
“Inferential stats use sample
data to try to make inferences
about a population which cannot
be known directly.”
(Glassman and Hadad, 2004)
Null Hypothesis

The null hypothesis always asserts that
only chance is at work


If a researcher can prove the null hypothesis
incorrect, it becomes more likely the
researcher is likely correct
Significance tests lead to probability,
ergo statistical inference is always a
matter of probabilities, never certainty.
Standard of Probability

Commonly accepted standard is 5 in 100


5 chances of being wrong makes 95 chances
of being right
Researchers print this value as p<0.05
Errors in Evaluation of Hypotheses
False positives (Type I) and false negatives
(Type II)


Experiment works but there is inconclusive
data to support=false positive (Reject null
hypothesis)
Experiment fails, but researcher overlooked
genuine effects=false negative (accept null
hypothesis)
There is no certainty is
inferential stats. This is why
there are no absolutes in
science. For most
psychologists, living with
uncertainty is part of the
challenge of understanding
behavior.