PSYCH 300

Midterm 1

Lecture 1

  • Producer role:

    • For coursework in psychology

    • For graduate school

    • For working in a university research lab

  • Jobs involving research:

    • Program evaluation

    • User experience

    • Marketing/advertising

    • Public health

  • Consumer role:

    • For psychology courses

    • When reading printed or online news stories based on research

    • For your future career: Evidence-based treatment

  • Jobs consuming research:

    • HR

    • Social worker

    • Education

    • Government

  • Empiricism (aka the empirical method or empirical research)

  • Quantitative methods: turn empirical observations into numbers

  • Qualitative methods: create rich descriptions not simplified into numbers

    • Reflexivity: Scientists self-consciously consider how their background or privilege shapes what questions they ask and what interpretations they make

  • Cupboard theory vs contact comfort theory

  • Theory is quite general, a more specific research question comes from it

  • Variables should be very clear from the research question

  • Studies don’t prove theories

  • Good theories are falsifiable

  • Scientists share the results of their research with the scientific community by submitting it to a scientific journal

Lecture 2

  • Experience has no comparison group

  • Should you stop hugging your dog? In 250 photos of people hugging their dogs, 82% of the dogs showed signs of stress (Coren, 2016). Which question would best help you evaluate the validity of the argument that “you should stop hugging your dog”?
        a. knowing how many dogs that were being hugged did not show stress
    signs
        b. knowing how many dogs that were not being hugged showed stress signs
        c. knowing what percentage of people usually hug their dogs
        d. knowing what the signs of stress are

    • Need to have a comparison group

  • Confound: an alternative explanation for some outcome other than the one we hypothesized

  • Behavioural research is probabilistic, which means that its findings are not expected to explain all the cases all the time (there are exceptions)

  • Sometimes, even good studies can yield “flukey” results that don’t represent reality

  • Responsible research producers strive to show that we can produce the same effect again and again (replicate)

  • Responsible research consumers pay attention to review articles and meta-analyses that summarize the results of many studies on the same topic

  • Ways intuition can be biased:

    • Being swayed by a good story

    • Being persuaded by what easily comes to minf

      • Availability heuristic

      • “If I can think of examples easily, it must be common/likely”

    • Focusing on the evidence we expect

      • Confirmation bias

      • “I already believe this, so I notice and remember evidence that supports it”

    • Being biased about being biased:

      • Bias blind spot

    • Availability heuristic and confirmation bias are NOT mutually exclusive

Lecture 3

  • Variable: something that changes or varies (within the context of the study/experiment)

  • Constant: something that stays the same (within the context of the study/experiment)

  • Measured variable: observed and recorded

  • Manipulated variable: controlled

    • Required for experiments (ie to demonstrate causality)

    • Some variables can only be measured and not manipulated

  • A frequency claim describes a particular level or degree of a single variable

    • Frequency claims involve only one measured variable

  • An association claim argues that one level of a variable is likely to be associated with a particular level of another variable

  • Association claims are supported by studies that have at least two measured variables

  • Variables that are associated are said to correlate

  • Can use associations to make predictions: the strong the association the better our prediction will be

  • A causal claim argues that one variable causes changes in the level of another variable

  • Causal claims are supported by experiments

  • Construct validity: Are we measuring the two constructs we intended to measure?

  • External validity/generalizability: Does our sample represent the population we are interested in?

  • Statistical validity: How precise is our estimate of the population frequency?

    • How strong is the association?

    • How precise/reliable is our estimate of the association?

  • Internal validity/Third-variable criterion: Can we be sure we’ve established causality?

  • Point estimate: A single estimate of some population value based on data from a sample

  • Confidence interval: A given range indicated by a lower and upper value that is very likely to contain the true population value for some point estimate

    • The CI gives us a range of plausible values

    • Narrower intervals indicate more precision

  • Random assignment reduces pre-existing differences between groups of participants

    • To make the groups similar, on average (before the manipulation)

  • Imagine we allowed children to choose what group they wanted

    • Group differences could be explained by the instructions

    • Or group differences could be explained by the different personalities of children in the groups

  • Prioritizing Validities:

    • Depends on the claim being made and the researcher’s priorities

      • Internal: Top priority for causal claims (not really in play for other claims)

      • External: Really important when you want to generalize to broader population/context

      • Construct: Pay extra attention to this when you’re examining something abstract

      • Statistical: Especially important when estimates are imprecise, effects are small, or when findings haven’t been replicated

Lecture 4

  • A research question states what the researcher wants to find out

  • A claim states what someone concludes based on their evidence

  • Three types of measures:

    • Self-report

    • Observational

    • Physiological

  • Nominal/Categorical Data: No inherent order to the categories

    • Thus, can only report how many people are in each category

    • NEVER calculate averages for nominal data

  • Ordinal Data: Levels in a clear order but not clear how much space there is between each level, can’t usually calculate average with this data

    • Job senority, birth order, place in a tournament, education level, preference of different foods

    • Levels = values that participants could choose from (or were assigned)

  • Interval Data: Even space between each level, can calculate averages

    • But no zero (or zero is not meaningful)

    • Temperature

    • Likert scales

  • Ratio Data: Even spacing between each level, zero is meaningful (even if it wouldn’t be picked by many/any participants)

    • Age

    • Number of years that you’ve worked with your current employer

    • Number of years of education

    • Speed

    • Reaction time

    • Number of times you’ve engaged in a behaviour (in a given time period)

    • Allows us to say things like “he is twice as old” ex

  • Reliability:

    • Test-retest reliability: Do we get the same result if we use the measure again (on the same thing)?

    • Interrater reliability: Do we get the same result if two people use the measure (on the same thing)?

    • Internal reliability: When a measure consists of multiple questions, do we get similar answers across all questions?

  • Correlation coefficient r: Direction indicated by whether it is + or -

    • Strength indicated by how close it is to -1 or +1

  • Scales: Sometimes we measure complex constructs (ex satisfaction with life) with multiple items (questions)

    • Participants indicate their agreement with each item using the same response options

    • Then average or sum each participant’s responses across all items

    • If designed well, responses to each item should correlate with (or be similar to) responses to the others

    • This indicates that the scale, as a whole, has high internal reliability

  • Internal reliability: Report “Average-inter-item correlation’ (AIC)

    • Looking for value >.15 (but less than .50)

    • or Cronbach’s alpha (more common in psychology)

      • Looking for something >.70

  • Subjective measures (face validity and content validity): Does it look like a good measure?

  • Statistical measures (criterion, convergent, and discriminant validity): Do the correlations make sense?

  • Both face validity and content validity are subjective ways to assess validity

    • Face validity: It looks like what you want to measure

    • Content validity: The measure contains all the parts that your theory says it should contain

  • Correlational evidence for criterion validity (outcome)

  • Another way to gather evidence for criterion validity is to use a known-groups paradigm

Lecture 5

  • Open-ended questions

  • Forced-choice format (best of 2+ options)

  • Semantic differential format

  • Things to avoid:

    • Double-barreled questions

    • Negatively worded questions

  • Shortcuts in survey responding:

    • Response sets (ex choosing 1 for every answer)

    • Acquiescence (yea-saying)

    • Fence sitting (choosing the middle/neutral option)

  • Observational data:

    • Observing how much people talk

    • Observing where babies and caregivers look

    • Observing sibling interactions

  • Observer bias: when observers see what they expect to see (observer’s INTERPRETATIONS of participants are influenced)

  • Observer effects: when participants confirm observer expectations (observer influences the ACTUAL BEHAVIOUR of participants)

    • Bright and full rats

    • Clever Hans

  • Masked (blind) design:

    • Observers are unaware of any information that could bias their observations OR convey expectations to participants

    • Typically, do not know what conditions participants have been assigned

    • Sometimes also unaware of the researcher’s hypotheses

  • Reactivity: change in behaviour when someone is aware they are being watched

    • Solutions: blend in (unobtrusive observations), wait it out, measure the behaviour’s results instead

    • Unobtrusive observation: observations made without participants knowing you are there

    • Participant observation: observations made while interacting with the people you are studying

Midterm 2

Chapter 2 (Stats)

  • Are Skittle flavours distributed equally in each bag?

    • Initial study: Conducted by Redditor Goobz24 with one party sized bag

    • Results visualized in a bar graph

    • Limitation: We can’t know from a single sample whether this reflects a real difference or is just sampling error (fluke)

    • Expanded study: Analyzed 36 jumbo bags using R software, found that across bags flavour distribution is fairly normal

      • But also that there is variable in both distribution of flavours and total number of Skittles per bag

  • Key purposes of graphs:

    • Summarize key statistical values

    • Visualize data distributions: all responses for a variable

    • Often we will start off with data in a frequency table

  • Frequency: Number of participants selecting a specific response

    • Relative frequency (or relative percent):

      • Frequency divided by the total number of participants (expressed as a percentage)

  • Frequency table components:

    • Rows: Response options (or categories)

    • Columns: Number (or %) of people who chose that option (or who are in that category)

  • Categorical vs Continuous data distribution

  • Highlight differences in average values or counts across experimental conditions or groups

  • Key elements of graphs:

    • Axes:

      • X-axis (horizontal): Labels the variable being measured

      • Y-axis (vertical): Displays values

      • Origin: Intersection of X and Y axes where both values are zero

    • Best practices for graphs:

      • Axis labels: Clearly define what each axis measures

      • Tick marks: Use consistent intervals for easy interpretation

      • Specific titles: Make the title descriptive

  • Pie charts:

    • Used to display categorical data

    • Show the proportion/percentage of observations in each category

      • Each “slice” represents one category, all slices together represent 100% of the observations

    • Best used when you have relatively few categories that make up a meaningful whole

    • Advantages: Good if you want to focus on proportions of a whole, don’t have to pick an order for our categories on the x-axis

    • Disadvantages: Can be hard to make precise comparisons between categories, can get messy if there are too many categories

  • Bar graphs:

    • Use bars to represent different groups/conditions

    • Examples: Y-axis shows raw frequencies, Y-axis shows percentages

    • Useful for comparing data across multiple categorical variables

    • Horizontal: Flipped axes, bars run horizontally instead of vertically

    • Stacked: Combine categories into one bar, with subcategories stacked

  • Histograms:

    • Similar to bar graphs but used for continuous variables rather than categorical variables

      • With continuous data, you can (theoretically) have participants who fall anywhere along a spectrum

    • Key features:

      • X-axis: Displays intervals or ranges of the measured variable

      • Y-axis: Represents frequency of the variable within each interval

  • Line graphs:

    • Visualize changes in a variable over time

    • Key features:

      • X-axis: Represents time intervals

      • Y-axis: Represents the unit of the measured variable

  • Scatterplots:

    • Use dots to illustrate relationships between two continuous variables

    • Key features:

      • X-axis: One continuous variable

      • Y-axis: Second continuous variable

    • Purpose: Identify and visualize relationships between variables

  • Density Plots:

    • Visualizes the distribution of a continuous variable

      • Wider areas: Higher frequency of responses (dense data)

      • Thinner areas: Lower frequency of responses

    • The violin (or cello, or bass) represents the entire distribution for one variable

  • Jitter Plots:

    • Visualize individual participant scores on a single quantitative variable

      • Each dot represents one participant’s response

    • Key features:

      • Axes: One variable measured on either X-axis or Y-axis

      • Clustering: Dense clusters of dots indicate popular responses

    • Shaped similar to a violin plot, but it shows the individual points

    • Advantages of using a jitter plot rather than a density plot:

      • Better if variable is not truly continuous

      • Can see how many data points there are

  • Normal distribution:

    • Symmetrical shape with upper and lower tails representing rare but possible cases

    • Also known as a normal curve or bell curve

    • Approximately half the data fall above and half below the mean (average)

  • Skewed distribution:

    • Data are not symmetrically distributed

      • Most observations cluster toward one end of the distribution, with a longer tail extending toward the other end

    • Types of skewness:

      • Positively skewed (right skewed): Data pile up on the left with a longer tail to the right

      • Negatively skewed (left skewed): Data pile up on the right with a longer tail to the left

  • Mode: The most frequent response for a given variable

    • Graphically, the peak of the data distribution

  • Unimodal data: One mode or peak in the dataset

  • Bimodal data: Two modes or peaks in the dataset

Chapter 3 (Stats)

  • Mean: Official statistical term for an “average”

    • Formula: M=∑X/n

    • M: Represents the mean

    • X: Collection of data points (ex survey responses)

    • ∑: Greek letter sigma, meaning “sum”

    • n: Sample size or number of data points

  • Mode: Most frequently given response for a variable

    • Best for summarizing nominal data (categorical, non-numeric)

    • Can also be used with datasets consisting of whole numbers

  • Median: The number that lies in the middle of all of the data points for a variable when the data are arranged in ascending order

    • Works well for interval or ratio data

    • Useful when data include outliers, as the median is not influenced by extreme values

    • Order data from smallest to largest: odd number of data points have the middle value as the median, for even number of data points calculate the averae of the two middle values

  • Use the mode to describe nominal data

  • Use the mean and median to describe interval or ratio data

  • Outliers: Extreme scores in a dataset that may skew results

    • Mean: Pulled toward outliers, affecting accuracy

    • Median: Resistant to outliers, representing the dataset’s true middle

    • (Mode also unaffected by outliers)

  • Symmetrical distribution: mean, median, mode are roughly equal

  • Variability: Describes the distribution of a variable

    • Describes the shape, distribution, or spread of data

    • Complements central tendency measures (mean, mode, median)

  • Range: A simple measure of variability showing the distance between the highest and lowest values

    • Formula: Range = Largest value - smallest value

    • Limitations: Does not give information about where the majority of the data points are

    • May overgeneralize

  • Standard deviation (SD): Typical distance of data points from the mean

    • Offers a more informative and precise measure of variability compared to the range

    • s=∑(X−M)2n−1s=\sqrt{\frac{∑\left(X-M\right)^2}{n-1}}

  • Variance: Square of the standard deviation

    • Intermediate step in calculating SD

    • s2 = ∑(X-M)2/n-1

  1. (X-M): Calculate how much each score differs from the mean

  2. (X-M)2: Square those differences (to give more weight to more extreme values)

  3. ∑(X-M)2/n-1: Sum and divide by n-1 (essentially calculating an “average difference” (but dividing by 1 less to make sure we don’t underestimate variability)


s=(X−M)2n−1s=\sqrt{\frac{\left(X-M\right)^2}{n-1}} : Square root the whole thing (to “undo” the square we did in step 2 and get back to the original units)

  • When data is distributed normally (bell shaped and symmetrical):

    • Approximately two thirds of the data is within 1 SD of the mean

    • 95% of the data is within 2 SD of the mean

    • 99.7% of the data is within 3 SD of the mean

  • SD helps us understand how far scores typically fall from the mean

    • Also helps us understand how a particular score compares to other scores in the distribution

Chapter 4 (Stats)

  • Z score (standardized score): refers to the number of standard deviations a given data point or observation is away from the mean for the variable

    • allows us to compare scores from two sets of data that may not:

      • have been measured on the same scale

      • have the same mean (average score)

      • have the same standard deviation (variability)

  • z = (X-M)/SD

    • X is a data point

    • M is the mean

    • SD is the standard deviation

    • The ratio of the difference between a single score and the mean to the “typical” differences of scores from the mean

    • How many SDs a given score is from its mean

  • Need to consider of z score is positive or negative:

    • Positive indicates the raw score is above the mean

    • Negative indicates the raw score is below the mean

  • Need to consider the size of the z score:

    • Absolute value: magnitude of a real number without regard to its sign

    • The higher the absolute value, the further the raw score is away from the mean

  • The normal curve is a distribution shape for 100% of the data for a variable, including the outliers

  • The entire area of the normal distribution curve (or any other curve) represents 100% of the data collected for a variable

  • The normal curve is symmetrical

  • We can figure out what portion of our data falls above or below any z score

  • Most of the data gather around the peak of the normal curve

    • Tails include the far right and far left portions of the extreme (less probable) data

  • Empirical rule: the portions of the normal curve that exist between z scores of -3, -2, -1, 0, 1, 2, and 3

    • When we standardize a score, we are seeing where it will fall on a distribution where the SD is 1 and mean is 0

  • Z scores can be used to understand the data distribution under the normal curve

  • Z table can be used to calculate different areas under the normal curve

  • Adverse impact refers to the rejection of job candidates based on their race, color, religion, nation of origin, and/or gender

    • These groups are considered protected classes who have historically faced hardships

  • Using z scores can reveal cutoff points for job applicants that may create an unintentional gender bias and/or other biases

  • Z score also reports the percentile, which refers to the percentage of individuals below z (larger portion of table)

    • Can calculate this by subtracting the “area beyond z” from 1

  • Probability: the likelihood of a given outcome

    • Usually presented as a percentage (ex 33%) or as a proportion/decimal (ex .33)

  • Remember when using the z table, scores below the mean will be -

  • To find raw score from z score, solve for X

  • The equal likelihood model is the model of probability with the premise that all outcomes are equally likely

  • Probability (p) of any outcome = 1/Total number of possible outcomes

  • The law of large numbers states that the expected distribution of equally likely events tends to become apparent only after the event has occurred a large number of times

  • Absolute risk (absolute chance): probability of an event occurring in a population

  • Relative risk (relative chance): probability of an event occurring in a group, relative to the same event occurring in another group

  • Conclusions based on relative values, though correct, can be misleading without the right context