1. Data
   1. Defn.
      1. A series of observations, measurements, or facts that can be analysed
   2. Variable
      1. Has a possible range of values
   3. Analysis
      1. Gathering, modelling and transforming data with the goal of highlighting useful information, suggesting conclusions and supporting decision making.
   4. Types
      1. Nominal
         1. Categorys
            1. No relationships
         2. Least powerful
      2. Ordinal
         1. Rank
         2. Has a relationship (1st, 2nd etc.
         3. Non mathematical relationship
      3. Interval
         1. No real 'ZERO'
            1. eg. temperature
         2. Has a mathematical relationship
      4. Ratio
         1. Has a tue 'ZERO'
         2. Eg. distance, height.
         3. Most powerful
2. Research Methods
   1. Which method?
      1. Depends on the question
   2. Quantitative
      1. Experiments
      2. Surveys
      3. RCTs
      4. Numbers
         1. Tells you what happened
   3. Qualitative
      1. Focus groups
      2. Interviews
      3. Case studies
      4. Tells you why it happened
   4. Validity
      1. Internal
         1. to do with study design
            1. Is it ok?
            2. Are we measuring the right thing
            3. Eg measuring height for as a measure of intelligence is wrong.
      2. External
         1. Can it be applied outside?
            1. Can the results be generalised?
   5. Replicability
      1. Can it be done again?
   6. Reliability
      1. If experiment done again will the same results be given?
      2. Easier for lab based work
   7. Objective
   8. Unbiased
   9. Variables
      1. Must be operational
         1. Be explicitly stated
   10. Constructs
       1. Defined by theoretical definitions
3. Variables
   1. Quasi-independent
      1. Characteristics that cant be randomly assigned
         1. Eg sex, age
   2. True experimental variables
      1. Can control these in a true experiment
      2. Can be randomly assigned
         1. eg Give Drug A or Drug B
   3. Independent variables
      1. The ones we control
         1. To bring about change in DV
      2. Levels
         1. At least TWO
            1. eg. gender- male/female
      3. AKA 'condition'
   4. Dependent variables
      1. The ones we measure
      2. The ones that depend on the IV
4. Research design
   1. Experimental design
      1. True experiment design
         1. Design where researchers can randomly assign participants to experimental condition.
         2. Eg randomly assign normal participants to consume different amounts of alcohol
         3. Randomised
      2. Quasi experimental design
         1. Design where researcher cant randomly assign participants to groups
         2. eG Compare heavy vs light drinkers, as you are either in one group or the other
   2. Randomisation
      1. Reduces confounding variables
         1. When groups to be compared differ in ways other than what the researcher has manipulated
         2. As they are distributed equally among the groups
      2. Prevents (un)intentional bias.
      3. Ensures participant is equally likely to be assigned to either group
      4. Enables use of powerful statistics.
   3. Subjects design
      1. Independent groups design
         1. Comparing BETWEEN groups
         2. Potential problems
            1. Confounding factors
            2. Solutions
            3. Randomisation
            4. Matched groups
         3. Matched groups
            1. Make sure subjects in both groups are matched as closely as possible on potential confounding factors
      2. Repeated measures design
         1. Testing WITHIN groups
         2. Advantages
            1. Fewer participant
            2. Each participant is their own control
            3. Removes some confounding factors
         3. Disadvantages
            1. Order effects
            2. Cant return ppnt to original state
            3. Practice effect
            4. better performance due to practice
            5. Fatigue effect
            6. SOLUTION
            7. Counterbalancing
            8. Randomly assigning order to group
            9. Therefore we can know whether the order has made any difference
   4. Causation
      1. How correct is our claim of A being the cause of B?
      2. SOLUTION
         1. Have a comparision group
            1. Eg treatment vs placebo
         2. Could do O-X-O
            1. Eg. test, give alcohol, test.
            2. Therefore we know if alcohol is the cause
5. Forms of validity
   1. Face
      1. Does it measure what it says it does?
   2. Criterion
      1. Concurrent
         1. Comparison of new test with established test
      2. Predictive
         1. Does the test have predictive value?
            1. Eg Does blood pressure value now predict heart attack in 5 years?
      3. Does the measured results agree with other measures of same thing?
   3. Construct
      1. How well does the design tap into the underlying construct
   4. Ecological
      1. Does study reflect naturally occuring behaviour?
         1. Eg does mouse in box reflect its behaviour in wild?
   5. Population
      1. Is our sample adequate for the claims we make about the population?
      2. What population are we interested in?
6. Sampling
   1. A sample is a selection or subset of individuals from the population
   2. Why sample?
      1. Time
      2. Money
      3. Sufficiency
         1. Maybe we dont need that much data as we feel that the sample gives an accurate data
      4. Access
   3. How
      1. Random sample
         1. No pattern
      2. Systematic
         1. Drawn from the population at fixed intervals
      3. Stratified
         1. Specified groups appear in numbers proportional to their size in population
      4. Opportunity/Convenience
         1. People who are easily available
         2. Leads to bia
      5. Snowball
         1. Get current participants to recruit more for the research
         2. Useful if you want to recruit very specific population
            1. eg drug users might know other drug users
7. Descriptive stats
   1. To describe a distribution we need to select the appropriate central tendency and distribution
   2. Central tendency
      1. Mean
         1. Average
         2. Less useful if there is a big outlier
         3. Best for continuous, symmetrical data
      2. Median
         1. Rank then find the middle value
         2. Best for ordinal data or interval/ratio data that is highly skewed
      3. Mode
         1. Most common
         2. Misleading if frequency is only just more than other values
         3. Best for nominal data
      4. For skewed data
         1. Positively skewed data, the mean has a higher value than the median, and the median has a higher value than the mode.
         2. Negatively skewed data, the mean has a lower value than the median, and the median has a lower value than the mode.
   3. Spread of data
      1. Range
         1. Max - Min
      2. Variance/Standard deviation
         1. Measure of mean deviation from mean
         2. SD
            1. The variability across individuals expected by chance
      3. Interquartile range
         1. 3rd quartile(75th qrtle)-1st quartile(25th qrtle)
         2. Useful when median is used as measure
         3. Value
            1. Large
            2. = data squashed
            3. Small
            4. = data spread out
         4. More stable than range as extreme values arent included
      4. Cumulative frequencies
         1. Each score and the number that attained that score and below
         2. eg if scores are 1-5, the we can say 7 people got 4 or less
      5. Percentiles
         1. Scores split into percentile
         2. A method of expressing a persons score relative to those of others
         3. Therefore if ur score is in 90th percentile you have done better than 90% of people
         4. 50th percentile=median
      6. Shapes of distribution
         1. Normal
            1. Ideal
            2. Symmetrical
            3. Mean/mode/median
            4. same
         2. Skewed
            1. Topic
         3. Kurtosis
            1. Steepness/flatness
            2. Steep
            3. Leptokurtic
            4. Flat
            5. Platykurtic
            6. +ve value= steep
            7. -ve value = flat
            8. 0=middling
   4. Z-score
      1. Converts a raw score into a number that shows how many standard deviations away it lies from the mean
      2. Allows us to see how different an individual is from the group.
      3. To calc. we need true mean+SD
         1. If unavailable, we have to use parametric tests.
8. Hypothesis testing+singnificance
   1. Probability
      1. The number of times the even of interest could happen divided by the total number of possible events
      2. If mutually exclusive
         1. Addition rule
      3. Sums up to 1
   2. P values
      1. Probability that you have rejected the null hypothesis when it was true.
      2. P<0.05
         1. Significant result
         2. Reject null hypothesis
   3. Types of error
      1. Type 1
         1. Incorrectly rejecting null hypothesis
      2. Type 2
         1. Incorrectly accepting the null hypothesis
   4. Types of hypothesis
      1. One tailed
         1. Difference in one direction
         2. eg eating sprouts increases your IQ
         3. Critical value
            1. Z=1.65
            2. If significant p-value is 0.05
            3. As getting a score outside this has a 5%chance
      2. Two tailed
         1. Difference can be in either direction
         2. Eating sprouts alters IQ
         3. Critical value
            1. +/- 1.96
            2. If significant p-value is 0.05
            3. As getting a score outside this has a 5%chance
9. Inferential stats
   1. Parametric tests
      1. Assumptions
         1. 1. Data normally distributed
         2. 2.Variance between groups is the same
         3. Therefore can only use data that conform with above
      2. T-test
         1. When SD+mean unknown
         2. Take a few measurements and estimate mean+variability
         3. #of measurements important
            1. Degrees of freedom
            2. #of measurements- #of parameters
            3. Higher the #, the more reliable the estimate.
            4. Closer the t-dist to the z-dist.
            5. Also decreases the critical value, as the fewer the measurements you take the larger the t-value has to be to reach significance
         4. Independent measures
            1. How to calculate
            2. Difference of means/difference expected by chance
            3. (Mean1-Mean2)/(SD of the means)
         5. Repeated measures
            1. Mean change/(change expected by chance)
            2. = mean difference/(standard error of difference)
         6. Shows us how different the means of the TWO groups are
      3. Standard error
         1. A measure of how close the sample mean is to the true mean
         2. Depends on SD of original distribution
         3. # of samples(n)
            1. The more samples you take the lower the standard error.
      4. ANOVA
         1. When more than two groups need to be compared
         2. Analysis of variance OF MEANS
         3. Between subjects design results in a variance between groups that is as a result of individual differences this means that the difference due to the factor being investigated has to be very large for the test to detect it.
            1. Therefore it is better to use WITHIN subjects design as this reduces the individual variability and therefore any difference due to the factor will be detected by the test.
            2. Therefore making the test more powerful
         4. F=Variability due to factor/(variability due to error)
            1. (mean btwn groups)2(sq)/mean within groups(2)(sq)
         5. Significance of F-value depends on two type of degrees of freedom
            1. 1.k-1 where k is #of groups
            2. 2.N-k where N is overall # of measurements.
   2. Non-parametric tests
      1. Sign test
         1. Weak
         2. REPEATED MEASURES ONLY
         3. 2 conditions
         4. At least 6 pairs
         5. Only shows direction not size of difference
         6. Only for DICHOTOMOUS DATA
         7. Method
            1. Disregard scores that stay same
            2. Count scores that go up or down
            3. The lower value is the calculated stat
            4. If this is smaller than the critical value then it is significant
      2. Wilcoxon test
         1. REPEATED EMASURES
         2. Method
            1. Rank data from smallest to largest
            2. Discard scores that remain the same
            3. Take away the smallest score from the largest for each person
            4. Rank the differences
            5. Add up the ranks for ppl who did best in condition A and Condition B separately
            6. The smaller value is 'T' the calculated stat
            7. Must be equal or LOWER than the critical value for conditions to be significantly different
         3. Takes into account the SIZE and the DIRECTION of the difference
            1. Therefore gives more info than sign test
      3. Mann-Whitney U test
         1. aka MAN-U
         2. For INDEPENDENT GROUPS
         3. Method
            1. Rank data as if it was one group
            2. Add up ranks for smallest group of both if groups are same.
            3. Take the smallest value (R)
            4. N1=#of cases in smallest group
            5. N2=#of cases in largest group
            6. U1&U2 are calculated
            7. Which ever is smallest is calculated stat
            8. U should be EQUAL OR LOWER than critical value for significance
      4. Kruskall Wallis test
         1. For INDEPENDENT GROUPS
         2. Ranking test
         3. Ranks within each group
         4. Take difference between mean rank of each group and total mean rank
         5. Square it & sum them up
         6. And that is used to calculate the statistic
            1. The larger the number the more likely that the conditions are significantly different
         7. Can tell you at least 2 groups are significantly different
            1. NOT WHICH TWO
            2. for that we need to draw it out
      5. Friedmans test
         1. For REPEATED MEASURES
         2. Method
            1. Rank within each individual's score
            2. Total up the ranks for each condition
            3. The computer measures the dispersion of the rank sums
            4. Looks at how different the total ranks are from eachother
         3. The Stat (S) has to be LARGER than the critical value to be significant