Transcript Ethics and Data Analysis

Research Ethics:
Ethics in psychological
research:
•
History of Ethics and Research – WWII, Nuremberg, UN, Human and
Animal rights
•
Today - Tri-Council (NSERC, SSHRC, CIHR)
•
Guidelines and Tutorial:
http://www.pre.ethics.gc.ca/english/tutorial/index.cfm
General Policy on Research Involving Human Subjects:
•
The researcher must inform participants about all aspects of the research
that are likely to influence their decision to participate in the study
•
Participants must have the freedom to say that they do not wish to
participate in a research project; they may also withdraw from the research
at any time without penalty
• The researcher must protect the participants from physical and
mental harm
• If deception is necessary, researchers must determine whether
its use is justifiable; participants must be told about any
deception after completing the study
• Information obtained on participants must be kept confidential
and researchers must be sensitive about invading the privacy of
the participants
Data Analysis:
• Topics:









Scales
Samples
Populations
Frequency Distributions
Measures of Central Tendency
Variability
Probability
Hypothesis testing
Significance
Scales:
• There are four basic types of scales:




Nominal
Ordinal
Interval
Ratio
Nominal:
• based on name alone
• Names or classes of
nominal variables
may have little if any
relation to one
another
Ordinal:
• based on order
• intervals between
units are not
necessarily equal
• (e.g. places of
individuals finishing a
race, 1st, 2nd, 3rd,…
are not separated by
equal time intervals)
Interval:
• intervals between
basic units on the
scale are equal
• has ordinal properties
• (e.g. degrees F,
degrees C)
Ratio:
• intervals between
basic units on the
scale are equal
• has ordinal properties
• has an absolute zero
(a value below which
others have no
meaning)
• (e.g. degrees K, all
weights and
measures)
Statistics:
• There are two fundamental types of statistics:
 Descriptive
 Inferential
• Descriptive: Used to summarize large sets of data
(e.g. correlations, frequency data, class averages
etc.)
• Inferential: Used to determine if experimental
treatments produce reliable effects or not
(inferences from sample to population)
Population:
• The entire group of
concern to a study
• Population data are
called parameters
Population
Sample:
• A subset of the entire
group of concern
• If a sample is derived by
random selection
– every member of the
population of concern has
an equal chance of being
selected for the sample
• Sample data are called
statistics
Population
Sample
Descriptive Statistics:
 Frequency Distributions
 Measures of Central Tendency
 Variability
Frequency Distributions:
• Tables, histograms, bar graphs, frequency
polygons, smooth curves
X
16
14
7
6
3
ƒ
2
4
6
3
1
Frequency Distribution Table
Histograms
Bar Graphs
Smooth Curves
Measures of Central Tendency:
• Estimate of where the majority of cases are in a
data set
• Mean: sum of all the individual datum divided by
the number of cases:
 For populations: µ and N
 For samples: M or X





n
• Mode: most
frequently occurring
score in a data set
• Median: middle most
score when data are
rank ordered
• Data:
7,6,8,6,8,6,6,6 (test scores)
• Rank order data:
6,6,6,6,6,7,8,8
 Mean = 6.625
 Median = 6
 Mode = 6
• So what do we mean by the term average ?
Relative position of mean, median and mode
with normal, positively and negatively skewed
distributions:
Normal Distribution
Positively Skewed Distributions:
Negatively Skewed
Distributions:
Variability:
• Variability refers to the concept of the
spread of a set of data
• Variability can be measured in several
different ways:





Range (largest number minus smallest)
Interquartile range
Semi interquartile range
Standard error of the mean (Inferential Stats)
Standard deviation (Descriptive Stats)
Standard Deviation:
• The average distance of scores in a data
set from the mean
 Calculating SD for a population
 Calculating SD for a sample
Calculating Standard Deviation for a Population:
X
(X-µ)
( X - µ )²
36
32
28
24
20
20
16
12
8
4
16
12
8
4
0
0
-4
-8
-12
-16
256
144
64
16
0
0
16
64
144
256
∑ X = 200
∑(X-µ)=0
σ
∑ ( X - µ )² = 960
µ=∑X/N
µ = 200/10 = 20
σ² = variance
σ² = ∑ ( X - µ )² / N
= 960 / 10
σ ² = 96
  ( x   ) 2 / N
 9.798
Calculating Standard Deviation for a Sample:
X
(X-X)
( X - X )²
36
32
28
24
20
20
16
12
8
4
16
12
8
4
0
0
-4
-8
-12
-16
256
144
64
16
0
0
16
64
144
256
∑ X = 200
S
∑ ( X - X ) = 0 ∑ ( X - X )² = 960
X=∑X/n
X = 200/10 = 20
s² = variance
s² = ∑ ( X - X )² / n - 1
= 960 / 9
= 106.67
s  ( x  x ) 2 / n  1
 10.33
Inferential Statistics:
• Based on hypothesis testing – making predictions
• Predicting whether sample effects will hold true at the
population level
• We can never be certain that effects seen at the sample
level hold true for the population
• Therefore we have to talk about the probability of an
effect in the population (given what is observed in a
sample)
• When conducting an experiment (using
samples) we create 2 opposing
hypothesis
• Working or Alternate hypothesis (H1):
 Drug X has an effect on the dependent
variable
• Null hypothesis (Ho):
 Drug X does not have an effect on the
dependent variable
• Basic procedure:
 attempt to disprove Ho. If this is possible, H1
is proven
 note: with sample data it is not possible to
prove H0, therefore, the hypothesis testing
procedure attempts to disprove H0
Example: Effects of a drug intended to
reduce symptoms of motion sickness:
• Hypothesis: Prediction of an effect
• Working Hypothesis: H1: Drug X helps
reduce the intensity of motion sickness
• Null Hypothesis: H0: Drug X has no effect
in reducing the intensity of motion
sickness
Significant effects:
• Significant means there’s a high probability of a sample effect being
true at the population level
• Significance, however, is expressed as the probability of our sample
effect being false at the population level (Type I error)
• The results of this study show that the drug significantly reduced the
symptoms of motion sickness (p < 0.05)
• p < 0.05 (minimum criterion for scientific publication)
• p < 0.01
• p < 0.001
• Note: Significance does not speak to the size of effects
Next class:
• Chapter 5: Development Through the
Lifespan