Research skills Quant Flashcards
What did Descartes and Locke, philosophers come up with
Descartes:
Rationalism - use of reason and logic to derive truth, sense deceive
Mind body dualism - both conceptually separate
Carteasian Dualism - mind and body conceptually separate but can interact
Locke:
Empiricism - knowledge of world constructed through experiences
Fechner and Wundt, how did impact experimental psychology
Fechner:
Developed psychophysics - uniting mind and body mathematically
Wundt:
Founding father of psychology
First psych lab at uni of Leipzig in 1879
Volkerpsychologie some things cannot be studied experimentally
Darwin and and Galton how helped methods
Darwin:
Proposed doctrine of natural selection
Led to study of individual differences - first scientific attempt to study emotions
Galton:
Argued for eugenics
Measured and classified human ability
Used intelligence tests, correlation, twin studies
What are the two types of reasoning
Inductive - incomplete
Bottom up approach
Reasoning from a singular statements to the probable validity of a conclusion
Deductive - complete
Top down approach
Reason from a general statement to a logical and certain conclusion
What did Karl Popper propose
Idea of falsification to test theories and hypotheses
Attempt to disprove theory then attempt to prove it
Can never be certain found final explanation
Why do we need research
Generate and test new ideas
Cannot rely on intuition, avoid myths
Develop objective evidence to inform knowledge
What is the:
Scientific method
Empirical method
Hypothetic-deductive method
Scientific method - general method of investigating using induction and deduction
Empirical: two stages
Gathering data, via experiences/sense
Induction of patterns and relations within data
Hypothetio-deductive method:
Creates hypothesis from observations
Develops theories
Test predictions from said theories
What is a null and alternative hypothesis
Null hypothesis - states no difference/effect/relationship between variables investigating
Alternative hypothesis - states there is a difference/effect/relationship between variables studying
Define generalisability and replication
Generalisability - extent to which findings can be generalised across a sample or population
Replication - repeating a research study in same way was originally conducted -
5 main quant data collection methods
Randomised controlled trials:
Randomly assigns participants to experimental or control group
Considered gold standard of research
Participants and researchers blind
True experiments:
Experimenter has full control on all variables
Experimental manipulation, standardised procedures, random allocation
Quasi experiments:
Like true but does not have complete control
No random allocation and/or no full control on IV
Correlational studies:
Used to determine if one factor is related to another
Study to what extent related
Non-manipulated variables
Observe natural variation and measure correlation
Questionnaires:
What is:
Independent variable
Dependent variable
Extraneous variable
IV - variable that experimenter manipulates as a basis for making predictions about DV
DV - variable that is measured or recorded in experiment
Extraneous - variables that potentially influence results but are not of direct interest to research
3 characteristics of variables
Continuous - can take any value within a given range.
Does not change in discrete jumps
Discrete - can only take certain discrete values within the range
Categorical - the value that the variable takes is a category
Within subjects design
How to deal with order effects
+ and -
Repeated measures - use same pp in every condition of the IV
Counterbalancing - ABBA, half do condition A first, other B. Spreads order effects across both
Elapsed time - leave enough time between conditions for learning or fatigue to pass
+ recruit less pp
+ equal groups
- order effects - caused by doing one condition then another, better with practice or fatigue
- demand characteristics - obvious aim when do twice
- attrition - lose one pp affects both conditions
Between subjects deigns
Considerations
+ and -
Independent measures - use different participants in each condition of IV
Use random allocation
Pre test - test skills or behaviour levels to match
+ no order effects
+ fewer demand characteristics
+ loss of pp only affects one condition not two
- if difference in variance of pp is too great limits statistical analysis
- participant variables from non-equivalent groups
What is a WIERD sample
Western
Educated
Industrialised
Rich
Democratic
4 types of probability based sampling
Systematic random sampling - every nth case from population
To make sure equal opportunity, randomly select starting point and chose every nth person from said point
Stratified random sampling - take random samples from various sub-selections of the population
Simple sampling - every member of target population has equal chance of being selected and all possible combinations can be drawn
Cluster samples - select clusters that represent sub categories. Groups in population samples at random from among similar groups and assumed to be representative of a population
3 types of non-probability sampling
Opportunity/convenience - whoever is available takes part
Self selecting/online - volunteer for research
Quota sampling - sample selected so that specified groups will appear in numbers proportional to their size in the target population
What is the history of ethics?
1947 after WW2, The Doctors Trail
Result the Nuremberg code was developed
10 ethical principles to protect participants in research
What 4 principles is BPS code of ethics and conduct based on
Respect
Competence
Responsibility
Integrity
What are the 4 levels of data
Nominal:
Categorical data with no particular over to rank importance
Ordinal:
Using scale/number to order/rank
Size between doesn’t mean anything
Interval:
Put scores in order
Equal difference between
No absolute zero
Ratio:
Same as ordinal
Has absolute zero
4 types of scales
Quasi interval - scale that appears to be interval but where equal intervals do not necessarily measure equal amounts of the construct
Ratio scales - interval type scale where proportions on scale are meaningful and has absolute zero
Discrete scales - not all subdivisions are meaningful, often where underlying constructs to be measured can only come in whole units
Continuous scale - no discrete steps, all points along scale are meaningful
3 measures of central tendency
Mean - sum of all scores, divided by number of scores in sample
Mode - most frequent score
Median - middle score when put in order
2 types of statistics
Descriptive - describe sample
Inferential - using what know from data to make inferences and generalisation from sample to wider population
What is:
Population mean
Sampling error
Sample statistic
Population parameter
Population mean - typical score in population
Sampling error - difference between sample statistics and population statistics
Sample statistic - statistical measure of a sample
Population parameter - statistical measure of a population
4 graphical descriptions of data
Bar chart - used to summarise categorical data
Separate bars as unrelated categories
Line chart - chart joining continuous data points in a single line
Histograms - type of bar chart for continuous variables
Bars joined, space shows no score in interval
Box analysis - exploratory data chart showing median, central spread of data and position of relative extremes
3 measures of variability
+ and -
Range - distance between lowest and highest score in sample
- sensitive to outliers
Interquartile range - distance between the upper and lower quartile in set of data
Semi interquartile range - half of the interquartile range
+ not affected by outliers
+ better then range, focusing on central units
- inaccurate with large class intervals
Standard deviation - estimate of the average deviation from scores from the mean
Indicator of how closely scores are clustered around the mean
+ most robust measure of dispersion
- sensitive to extreme values
What are the two ways to calculate SD
Corrected - used to estimate population standard deviation
Uncorrected - used when not using the standard deviation to make estimates of the underlying population
Characteristics of normal distribution
Peak in middle
Bell shaped curve
Tails off symmetrically at either side of peak
Generally more scores plotted, more like normal distribution becomes
What is positive and negative skewed distribution
Positive - peak shifted to left, tail towards right
Negative - peak shifted to right, tail towards left
Define bimodal and multimodal distribution
Bi modal - 2 peaks in distribution
Multi modal - >2 major peaks in distribution
What is:
Kurtosis
Leptokurtic
Platykurtic
Mesokurtic
Kurtosis - measure of peak and flatness or steep and shallowness
Leptokurtic - higher kurtosis/very peaked distribution
Platykurtic - lower kurtosis/flat distribution
Mesokurtic - between two extremes of peak and flatness
What is the standard normal distribution
Distribution of z scores
What is a z score
How do you calculate
how many standard deviations above or below a mean score is
Subtract the sample mean from the score
Then divide by the sample standard deviation
What is:
Sample mean
Population mean
Sample mean = mean of sample - subset of population
Population mean = mean in population
What is confidence intervals
How do calculate confidence intervals
Probability that a population parameter will fall between a set of values
95% CI used
Based on 2SD (1.96)
1.96 SD above and below mean = 95% of the SND
So 95% confident our sample will be within 1.96 SD of the population mean
How accurately our data reflect the true population is dependent on the standard error
Define these hypothesis:
Directional
Non-directional
Causal
Non-causal
One-tailed
Two-tailed
Directional - suggests direction of effect
Non-directional - does not specify difference/effect
Causal - suggests casual inference
Non-causal - suggests specific characteristics of behaviour without reference to behaviour
One tailed - have specified direction of relationship between variables
Two tailed - have predicted that will be a difference but not prediction direction
What are Cohen’s effect size guidelines
Small - D = 0.2
Medium - D = 0.5
Large - D = 0.8
What is the P value
Probability of observing the effect as large as observed or larger, if the null hypothesis is true
Shows how likely it is that your data could have occurred under the null hypothesis
At what P value should the null hypothesis be rejected
P<0.05
The smaller the p value, more likely to reject
What are key differences between
Parametric and Non-Parametric tests
Parametric:
Based on population parameters
Assumptions about the underlying population our data is from
More assumptions
Less universal
Distributed data
Larger power
Non-Parametric:
No strict assumptions about the data distribution
Can be used when assumptions are met and not met
More universal
Continuous data
Lower power
4 parametric assumptions that would mean a non-parametric test would be needed
Scale which we measure DV should be interval or ratio level
Populations the sample is drawn from should be normally distributed
Variances of the populations should be approximately equal if comparing more then one group
No outliers or extreme scores
What is T-test
When to use
Devised by William Gosset statistician working for Gyuiness
Developed idea of how to make inferences about small differences in population based on differences between small samples
Asses how likelihood of obtaining the observed differences between two groups by sampling error
Used when want to compare differences in means:
Two separate groups
One group measured on two occasions
Wether one group differs from a specific mean
Parametric test - populations the samples drawn from should be normally distributed
What is degrees of freedom
How is it calculated
Number of individual scores that can vary without changing the sample mean
Number of observations made - number of parameters established
What are the df for:
One sample t test
Related t test
Unrelated t test
One sample t test - N-1
Related t test - N-1
Unrelated t test - (N-1)(N+1)
Difference between repeated/within subjects t-test and independent/between subjects t-test
Repeated/within subjects t-test = used when exploring differences in a within groups design using same participants
Independent/between subjects t-test = used when exploring differences between subjects using different participants
What is the correlation coefficient
How does it relate to covariance
How do you find shared variance
Strength of relationship between two variables (r)
+1 = perfect positive relationship
0 = no liner relationship
-1 = perfect negative relationship
Correlation coefficient is ration between covariance and a measure of separate variance
When two variables correlated, share variance
Square the correlation = get shared variance
What is Cohens effect size guidelines
Small - R - 0.1
Medium - R - 0.3
Large - R - 0.5
4 steps in formally reporting statistical results
State type of correlation performed, variables correlated, state direction of relationship found
Report the test statistic, df, statistical significant
Report effect size
Comment on direction of relationship
What is the third variable problem
How to handle
Other measured or unmeasured variables that affect results
Partial correlation calculates what the relationship between two variables would be if take away the influence of additional variables
5 things parametric tests assume
Underlying probability distributions ie normal distributions
DV measured at interval or ratio level
No outliers
Homogeneity of variances
Linearity
When not met: data may be non-normal, not at required level, outliers, sample size too small, unequal sample sizes if using groups
What 3 tests are parametric
What 3 tests are non-parametric
Parametric:
Independent samples t test
Related samples t test
Pearsons product moment correlation
Non-parametric:
Mann whitney u
Wilcoxon signed rank
Spearman’s rho
What 3 non-parametric tests are alternatives for parametric tests
Mann whitney u = alternative to independent samples t - test
Wilcoxon signed rank = alternative for related samples t test
Spearman’s rho = alternative to pearsons product moment
What is internal validity
Threats to
How to improve
Extent to which an effect found in study was caused by manipulation of IV
Attrition - pp dropping out
History - events between measurements
Sampling -
Maturation - pp changing over course
Testing and instrument issues - repeating ie order effects
Standardised procedures - for both researcher and pp
Counterbalancing - avoid order effects
Single or double blinding - eliminate research expectations
What is external validity
Degree to which results generalise beyond the experimental context
What is external and internal reliability
External:
Test-retest ability - correlation of peoples scores at one time and same at later
Inter observer or inter rater reliability - ability to which researchers agree in their ratings or codings
Internal:
Internal consistency of test
Participants scoring similarly across multiple items of a construct
Most commonly used to measure = Cronbach’s alpha - calculation of how closely related a set of items are as a group
