Experimental Methods Exam #2
You hypothesize that exercise will decrease stress. In particular, you believe that different types of exercise may be more effective in reducing stress than others. Thirty participants were randomly assigned to three groups of 10 to study the effects of different forms of exercise (basketball; jogging; or swimming) on reducing stress. Each participant was given a stress test at the end of the semester.
(1 x 3) One-way between subjects ANOVA
In Detroit, more red cars are involved in accidents in the summer than in the winter. You hypothesize that when people are hot, their driving skills decrease. You decide to test six groups of participants on a motor skills test. The effects of temperature (70, 80, or 90 degrees) and oncoming car color (red vs. blue) were tested on a 20-min driving test.
(2 x 3) Between subjects factorial ANOVA
You are interested in the effects of music training on music perception. To test your ideas you recruit 10 trained musicians and 10 untrained musicians. Next, each participant listened to three 10-min musical pieces (a highly discordant composition, a highly harmonious composition, and a piece that alternated discordant and harmonious passages). After each musical piece, all participants rated the piece on its complexity and its listenability. Which statistical technique would you use to determine if the effects of music training on the perception of music complexity in music?
(2 x 3) Mixed Factorial ANOVA
You have a new anti-viral drug that you think will help AIDS patients. The antiviral drug was presented in one of four concentrations (.01cc, .025 cc, .05 cc., and .10 cc). The patients either had AIDS or were controls without AIDS. Following a 2-week drug regimen, the patients' white blood cell counts were measured.
(4 x 2) Between subjects factorial ANOVA
You notice that you have a harder time reading and concentrating when it is noisy. A group of participants read a 5-page passage for 15 min in each of five different noise conditions. The different noise conditions were: 1) 0 Hz = no noise; 2) 25 Hz; 3) 50 Hz; 4) 75 Hz; and 5) 100 Hz. The participants completed a 20-queston test on each reading.
(5 x 1) Within subjects ANOVA
You are interested in the effects of college coursework on women's leadership goals. Specifically, you have learned that women's leadership aspirations are high when they enter college, but they drop dramatically by the time they graduate. To determine if and when these changes occur, you recruit 100 female first-year students at a small liberal arts college. You administer a valid measure of leadership aspirations at the beginning of their first year, sophomore year, junior year, senior year, and at graduation. Which statistical technique would you use to examine changes in women's leadership goals?
(5 x 1) within subjects ANOVA
Non-Experimental Designs Nonreactive/Unobtrusive Measures: Physical Traces (2 Types)
*A source of evidence that is based on remnants, fragments, and products of past behavior* 1) Use Traces 2) Products
Mixed Factorial ANOVA
*BOTH BETWEEN GROUP AND WITHIN GROUP CONDITIONS* *TWO OR MORE IVS*
Non-Experimental Designs Quasi-Experimental Designs (5 types)
*O's represent observations, X's represent treatments 1) Time Series Designs 2) Interrupted Time Series Design 3)Multiple Times Series Design 4) Non-equivalent Before-After Design 5) Ex Post Facto Design
Non-Experimental Designs Quasi-Experimental Designs: Time Series Designs
*O's represent observations, X's represent treatments treatment: ( O1 x O2) *Many confounds, life goes on and it's impossible to control for
Between subjects (FACTORIAL) ANOVA
*TWO MORE MORE IVS (MULTIPLE IVS)* -Working with diff groups
Counterbalancing in Repeated Measures Incomplete Designs (2 Types)
*each participant will experience each condition only onece 1. Random Starting Order -e.g., CNEI, NEIC, EICN, ICNE 2. All Possible Orders
Non-Experimental Designs Nonreactive/Unobtrusive Measures: Physical Traces - USE TRACES (2 Types of Measures)
*physical evidence resulting from use (or nonuse) of an item (checking trash cans to see their recycling behavior)* 1) *Accretion Measures*: based on the accumulation of material (underlined book) 2) *Erosion Measures*: based on selective wear (how much has that been used) e.g., Friedman & Wilson, 1975-textbook study: how much do students use textbook? Experimenters went to two pages and glued corners and then saw at the end of the term when students returned books how many students tore the pages to read the assigned chapter.
Counterbalancing in Repeated Measures Complete Designs
*sees conditions more than once 1. ABBA Counterbalancing -All conditions experience twice in different order -e.g., ABC, CBA -*Problem* with this is *anticipation effect* 2. Block Randomization -n x n array with each condition appearing at least once in each ordinal position -gets all four conditions -order of blocks would be counterbalanced -goes in different order *kind of like Latin square?
Types of Factorial ANOVA Interactions (4 Different Interactions) "Divergent" Interaction (Characteristics; main effect)
- lines in *opposite directions* -*CHARACTERISTICS* 1) The effect of A1 vs. A2 is stronger at B2 than it is at B1 (height of black bar (A1) vs height of light bar (A2)) 2) The effect of B1 vs. B2 is opposite at A1 from what it is at A2 (slope of black bar (A1) vs. slope of light bar (A2)) *The dark line (A1) is always better than the light line (depending on pic, it's whichever one is on top I think) *MAIN EFFECT* of either A or B, not both. The effect of A1 vs A2 is stronger at B1 than B2 (separating)
Confidence Intervals (CI)
-*A range of possible values in which we feel confident that the ACTUAL GROUP value will lie in the larger population* -Most common CI is 95%, but any value could be acceptable (e.g., 90%, 99%) -aka if we test infinite # of samples and calculate internal for each sample, 95% of those sample would contain a mean/mean difference within this range
Non-Experimental Designs Nonreactive/Unobtrusive Measures: Archival Data
-*Archives*: Records/documents recounting activities of individual, institutions, government, and other groups. -obtained by inspecting the records and documents produced by society and by analyzing media reports
Types of Factorial ANOVA Interactions (4 Different Interactions) "Crossover" Interaction (Characteristics; main effect)
-*CHARACTERISTICS* 1) The effect of (height of black (A1) vs light bar (A2)) A1 vs A2 is the opposite at B2 from what it is at B1 2) The effect of B1 vs B2 is opposite at A1 from what it is at A2 *No MAIN EFFECT of A or B (both B and A are in same spots)* -look at average mean of each variable
T-test
-*difference* -IQ, number of words recalled, rating on a scale -equal intervals between points -ratio/intervals (continuous)
Pearson's r
-*relationship* -IQ, number of words recalled, rating on a scale -equal intervals between points -ratio/intervals (continuous)
Non-Experimental Designs Single-Participant Designs: Experimental Single Participant Designs - REVERSAL DESIGNS
-ABAB Reversal Design "water faucet effect" -Burns & Swerdlow, 2003: normal and then liked pedophilia behavior; had tumor in brain that did this. took tumor out and thoughts stooped. Tumor later grew back and thoughts came back too. -A is observation, B active stage is going on, then go back to A without treatment (B), and then add treatment (B) again -e.g., cussing and posting it on fb -a-passive design, b- active stage (go and stop of results)
Post hoc tests
-After the fact, do tests after seeing "F" score -To explore a *sig. one-way ANOVA*, you use a post hoc *(pairwise comparison)* -Ex. Student-Newman-Kewls; Tukeys
Analysis of Variance (ANOVA)
-All ANOVAs calculate F Ratio
Meta-Analysis
-An analytical technique in which you combine the results from multiple published reports to determine the occurrence of a phenomenon -in particular, ES and CI are invaluable when comparing across different published reports.
Effect Size (ES)
-Be able to understand them if you encounter them on EXAM -*The amount of variability in a DV that can be traced to the IV (i.e., effect size provides an estimate of the strength of the effect of the IV on the DV)* -A better measure of *MEANINGFULNESS*
Non-Experimental Designs Quasi-Experimental Designs: Multiple Time Series Design
-Control for natural changes in comparison with no treatment and treatment O1 O2 O3 x O4 O5 O6 O1 O2 O3 O4 O5 O6 (control groups)
Non-Experimental Designs Nonreactive/Unobtrusive Measures: Physical Traces - Products
-Creations, constructions, or other artifacts of earlier behavior -e.g., "suicidal poet study (Stirman & Pennebaker, 2004): more introspective language and words in poets who committed suicide
Non-Experimental Designs Single-Participant Designs: Cons (Disadvantages)
-Difficultly of drawing cause - and - effect conclusions -Sources of bias interpretation or collection (e.g., patient talking about incident with mother - therapist doesn't know mother so they can't verify these incidents (*faulty memory*) -Generalization Problems -very difficult to generalize from one individual and apply it to general population
Non-Experimental Designs Single-Participant Designs: Experimental Single-Participant Designs (Two Types)
-Dolphin ex. (way to communicate with and give sentence structures to dolphins) A. Reversal Designs (ABAB) B. Multiple Baseline Designs
Factorial ANOVAs Main Effects
-F-score for main effects tests whether there is *an overall effect of that factor ignoring the other factor/factors* -Compares *marginal means* (i.e., row/column means) -row or column means (means of whole row to each other or mean of whole column to each other
Describing Factorial ANOVA Factorial ANOVAs and Designs
-Factorial ANOVAs and Designs are described as a *matrix* with particular dimensions (e.g., 3 x 3, 2 x 2, 4 x 3 x 2) *each number represents an IV* -The values of each number represents the *levels that IV has*
Concerns with Null Hypothesis Significance Testing (NHST)
-G. Cumming (2014) "The New Statistics" -"Chasing p-values"
Non-Experimental Designs Nonreactive/Unobtrusive Measures: Archival Data - TYPES
-Media Reports -Running Records (diaries) -Ex. Nun Study: Snowden: stroke and head trauma can boost chances of coming down with illness like Alzheimer's later in life. College education and active intellectual life may protect you from disease. How you use your brain can influence how your brain responds to these disease later in life
Non-Experimental Designs Quasi-Experimental Designs: Interrupted Time Series Design
-Multiple observations before and after treatment (O1 O2 O3 x O4 O5 O6) -Less confounds, control for natural changes in addition to the treatment -*within participant*
Quasi-Experimental Designs
-Open systems vs. closed systems -Open systems: may have little control (e.g., schools, sport systems) -Closed systems: experimenter has control over situation, cause and effect available (opposite, experimental) -Used when sufficient control over the variables under study and potentially confounding variables are lacking. Because of this lack of control, definite statements cannot be made about cause-and-effect relationships -trade off between external and internal validity
Follow Up Tests for Repeated Measures ANOVA
-RM ANOVA formula will generate a single F score, and larger F scores are indicative of statistically sig. group diff.
Non-Experimental Designs Single-Participant Designs: Pros (Advantages)
-Rare phenomena (e.g., feral children) -Idea Generation -Challenge to *theoretical assumptions* (e.g., "all swans are white" -one piece of evidence to ruin assumption (value of a case study) -Clinical innovations
Factorial ANOVA INTERACTIONS
-The F-score for an INTERACTION is testing whether the effect of one IV *depends on the other IV* -If the effect of one IV is constant across the levels of the other IV, there is *no interaction* -Interactions can exist *alongside main effects* or *all on their own*
Partial Eta Squared (Np^2) (ANOVA)
-Value tells you the percentage of *variability* in the DV that is accounted for by the IV -so if Np^2 is 0.17, it means that 17% of variability is explained by the IV -To date there are no established guidelines for classifying a Np^2 as weak, moderate, or strong (but *bigger is better*) -*used for ANOVAs* *no ceiling limit
Non-Experimental Designs Single-Participant Designs: Descriptive Single-Participant Designs
-case studies -*extremely long analysis of individuals*
Interpreting Cohen's d
-d < 0.20 = trivial effect (92% overlap) -0.20 < d < 0.50 = weak effect (80% overlap) -0.51 < d < 0.80 = moderate effect (69% overlap) -0.80 < d = strong effect *for t tests* *no ceiling limit
Non-Experimental Designs Single-Participant Designs: Experimental Single Participant Designs - MULTIPLE BASELINE DESIGN
-designed to gain some measure of experimental control when studying one participant. Here, one simultaneously monitors several behaviors of single participant -Multiple baseline across *BEHAVIORS* -Multiple baseline across *SUBJECTS* (e.g., Baahrick et al., 1993: words that were most spaced out were ones most remembered years late) -Multiple baseline across *SITUATIONS* (can we get behavior to change across multiple contexts.
Chi-Square
-e.g., gender, frequency of people who prefer coke to pepsi -Frequency data -Nominal/categorical
Types of Factorial ANOVA Interactions (4 Different Interactions) "Synergistic" Interaction (Characteristics; main effect)
-lines in *same direction*, one is just *more steep* -*CHARACTERISTICS* 1) The effect of A1 vs A2 is stronger at B2 than it is at B1 2) The effect of B1 vs. B2 is stronger at A1 than it is at A2 3) The combination of A1 and B2 produces bigger effect than what you'd predict from combinations of individual effects *MAIN EFFECT* of A or B. The effect of B1 v B2 is stronger at a1 than A2 (angled separation) -having both a and b produces bigger effect than predicted of individual effects
ANOVAs and F-scores
-mean square/error = F -F score for each variable *Factorial ANOVAs: will always have 3 F scores, and you have to report at least 3* Factorial ANOVA with 2 IVS: 3 F scores -Factor A, Factor B, A x B Factorial ANOVA with 3 IVS: 7 F scores -Factor A, Factor B, Factor C, A x B, A x C, C x B, A x B x C
Non-Experimental Designs Single-Participant Designs
-n = 1 designs -case studies -focuses on single individual
Non-Experimental Designs Quasi-Experimental Designs: Ex Post Facto Design
-natural treatments -after the fact (X O1 O2 O3) ex. tsunami, recovery from traumatic events
Non-Experimental Designs Quasi-Experimental Designs: Non-equivalent Before-After Design
-no random assignment -predetermined groups O1 x O2 O2-O1* O1 O2 O2-O1 (control group) *before - changes in improvements between the groups -groups are already different so you have to include all
Problems with NHST
-prompts us to think dichotomously -conflation of "significant" with "important" -just because something is significant doesn't mean it's important; just tells us groups are different -p values are susceptible to sample size -p values tend to vary dramatically from one replication of an experiment to another
Spearman's rho
-ranked data -ranked preferences -ordinal
Types of Factorial ANOVA Interactions (4 Different Interactions) No Interaction (Characteristics; main effect)
-whenever you have *parallel lines*, there is NO interaction -*CHARACTERISTICS* 1) The effect of A1 vs A2 is constant at every level of B (i.e., A1 is always better than A2 by same amount at both B1 and B2) 2) The effect of B1 vs B2 is constant at every level of A ( think about slope of a line) (i.e., B2 is always better than B1 by same amount at both A1 and A2) -you would expect *MAIN EFFECT* of A because A1 and A2 are not on top of each other (A1 greater than A2) -*MAIN EFFECT* of B1 w/ B2 greater than B1
Repeated Measures ANOVA
-will have at least 2 IVs
Three Types of Factorial ANOVAs
1) Between Subjects -each cell has diff group of people (only one combo) 2) Repeated Measures (within subjects) -every cell contains the same people (does all IV) 3) Mixed -participants contribute to all levels of IV but only one of the other (mix between w/in and b/w)
Types of Effect Size
1) Cohen's d (t-test) 2) Partial Eta Squared (Np^2) (ANOVA) *helps you understand how strong of a significance there was - verify the significance* *follow up tests*
Types of Factorial ANOVA Interactions (4 Different Interactions)
1) No Interaction 2) Crossover Interaction 3) Divergent Interaction 4) Synergistic Interaction
Four Types of Analysis of Variances (ANOVAs)
1) One-way (between subjects) ANOVA 2) Repeated Measures (One-way Within-subjects) ANOVA 3) (Between subjects) Factorial ANOVA 4) Mixed (Factorial) ANOVA
Non-Experimental Designs Nonreactive/Unobtrusive Measures (2 Types)
1) Physical Traces 2) Archival Data *observations that are not influenced by the presence of the investigator*
Follow Up Tests (For ANOVAs in General)
1) Planned Comparisons 2) Post hoc tests
Non-Experimental Designs (3 types)
1) Quasi-Experimental Designs 2) Single-Participant (n=1 designs) 3) Nonreactive/Unobtrusive Measures
Basic Components of a Table
1) Table Number 2) Table Title 3) Column Headings 4) Row Items 5) Underneath table, you need table notes *table entries that are meant to be compared should be near one another*
Non-Experimental Designs Single-Participant Designs: Problems of Experimental Single Participant Designs
1) Variability on behavior of organism 2) Cannot use inferential statistics 3) Cannot study interactions *
Writing Results and Conclusions from Factorial ANOVA Important Things to Include
1) specific descriptions of the type of statistics reported 2) means and standard deviations of each group 3) the statistical summary statement for all main effects and interactions (YOU SHOULD INCLUDE AT LEAST 3 F SCORES) 4) Appropriate planned comparisons or post hoc tests for significant effects 5) Interpretation of findings 6) Reference to figure *order of things should flow well *you can never have an interaction between an IV and DV* -*you have an interaction on an IV and IV on a DV*
Non-Experimental Designs Nonreactive/Unobtrusive Measures: Archival Data - PROBLEMS
1. *Selective Deposit*: When biases exist in production of archival source (things don't get recorded) 2. *Selective Survival*: records are missing or incomplete 3. *Spurious Relationships*: not controlling variables. -e.g., marriage rate in Alabama correlates with number of people who were electrocuted by phone lines by year (r = .9)
Describing *Factorial* ANOVA Example: Researchers measured the average number of hallucinations participants reported after taking either a low or high dosage of one of two drugs: LSD or marijuana. They also compared participants with a past history of drug use to those with no history of drug use. 1. What is the *design* of the study? 2. How many groups of participants are there?
1. 2(Dosage: Low vs. High) x 2(Drugs: LSD vs. Marijuana) x 2(History of Drug Usage: Yes vs. No) 2. 8 groups of participants (2 x 2 x 2)
Concerns with Null Hypothesis Significance Testing (NHST) 3 Recommendations from Cumming
1. Abandon NHST (i.e., p values) 2. Always report effect sizes and confidence intervals 3. Conduct more meta-analysis
F Ratio
= systematic variance/nonsystematic variance -Systematic variance reflects differences between your groups from your IV
Factorial ANOVA Example One group wanted to explore the perception of sexualized humor. They did so by having male and female participants rate the level of amusement and the level of offensiveness to a dumb blonde joke. Moreover, they wanted to determine if the amusement or offensiveness was affected depending on the sex of the joke teller. To accomplish this experiment, they videotaped a scripted conversation between either 2 males or 2 females. During this brief conversation, one of the speakers included a dumb, blonde joke; the joke (and all other aspects of the conversation) was the same regardless of the sex of the actor. Once the participants had completed viewing the videotape, they completed a number of questions, including a rating of level of amusement of the joke and a rating of level of offensiveness of the joke on a 7-point Likert scale with low values indicating less agreement and high values indicating more agreement. What type of design did the group employ?
A 2(Sex Participant: Male vs. Female) x 2(Sex Speaker: Male vs. Female) between subjects factorial ANOVA *For both amusement and offensiveness
Factorial ANOVA Example One group wanted to expand on the t-test experiment in which males and females tend to remember words within their gender significantly more than words outside of their gender. Their unique hypothesis was whether this would differ depending on the sexual orientation within the genders. Of course, we discussed that sexual orientation and gender are not exactly the same variable, and just because an individual has a different sexual orientation does not mean they have a different gender role orientation. They said they were aware of this, and planned to use Bem's Sex Roles to make these adjustments. Their design included the subject variable of the 5 categories on Bem's Sex Role scale, and the within-subject variable of word type to include neutral, masculine, and feminine words. Identify the type of design they created and how many groups they would need to test their hypothesis?
A 2(Sexual Orientation: Heterosexual vs Homosexual) x 2(Gender: Male vs Female) x 5(Bem's Sex Role Scale) x 3(Word Type: Masculine vs. Feminine vs. Neutral) between subjects ANOVA (20 groups of people - 2 x 2 x 5) - 3 (word type) is a within subject factor so it does not need to be included (I think)
F Score
A sig. F score means only that some sig. differences exist between your groups/conditions -Stat that goes with ANOVA
Counterbalancing in Repeated Measures
A. Incomplete Designs B. Complete Designs
Concerns with Null Hypothesis Significance Testing (NHST) APA's Current Approach
Author(s) should provide: -p values to test null hypothesis -Effect sizes -Confidence Intervals
You have a new drug that you think will help AIDS patients. A group of 20 AIDS patients are randomly assigned to two groups of 10. One group is treated with a new drug, AXTZ, and the second group is given a placebo drug. The white blood cell count of the participants was tested 24 h after dug administration.
Between subjects independent t-test (independent samples t-test)
Planned Comparisons
Comparison plan to make before experiment
P-hacking
Data analysis and reporting are often selective and biased -In many research fields, studies are rarely replicated so false conclusions persist. (*what I wanted to go back and do for Project 2 when nothing was significant*)
Factorial ANOVA and Design
Factorial ANOVA allows us to systematically calculate effect of *each factor (independent variables)* separately, as well as *their interaction*
For Project 2 in the Experimental Methods class, your group wants to investigate the effects of social exclusion on motivation for future social interaction. using random assignment, you place participants in two groups, a social exclusion group and a control group. Participants in the social exclusion group are given a passage to read designed to make them feel socially excluded while control group participants read a neutral passage. You then ask all participants to answer questions on a Likert scale about their desire to participant in a hypothetical student group. How would you determine whether there are statistically significant differences between the desire of participants in each group to participate?
Independent samples t-test
The owners of Starbucks recently claimed that college students who drink coffee daily have higher GPAs than those who do not. You have been looking for a way to justify your coffee addiction and decide to survey the Kalamazoo College campus using surveymokey online, asking people to report whether or not they drink coffee on a daily basis and their GPA. What statistical test should you use to analyze your data?
Independent samples t-test
You hypothesize that wearing pink clothes makes people happier than wearing black clothes. You give on group of participants pink velvet track suits and another group black velvet track suits to wear for a day. At the end of the day, you ask participants to rate their happiness on a 1-7 scale. Which statistic would you use to determine if the color of people's track suits affected their happiness?
Independent samples t-tests (interval)
Factorial ANOVA Example Looking at a bar graph with males and females who like expensive or cheap advertisements: Men liked the cheap ad, women like the expensive ad What kind of interaction is this?
It is a CROSSOVER interaction, because each group had *flipped* preferences -If the preference is flipping/flipped between groups then it is a crossover interaction
Practice Problems You hypothesize that cows with more black spots eat more grass. You decide to count the number of spots on each cow in your herd and measure how much grass each cow eats. Which statistic would you use to determine if a relationship exists between number of spots and amount of grass eaten?
Pearson's r
Studies have investigated sexual arousal during distraction in both men and women. Usually the participants are in a laboratory setting and are presented with an erotic video or audio clip. A distracting task in another mode (if the clip is visual the distraction will be audio) is given and the differences are assessed. For women, assessment of physiological sexual arousal in the lab is measured using a vaginal photoplethysmograph, a tampon size probe (inserted by the woman herself) that measures vaginal pulse amplitude (VPA), or increased blood flow. In a proposed study, women listened to an audio clip of an erotic episode from a female perspective. During the clip they were shown a series of numbers and were asked to add subsequent pairs of digits and verbally report their sum, this was their distracting task. The women also gave a subjective sexual arousal rating on a 7-point Likert Scale every 2 minutes. What statistic would the researchers use to see if physiological and subjective sexual arousal were related across participants?
Pearson's r
A group of students realized that when they slept fewer hours because of stress, they were more likely to drink more alcohol on the weekends. In order to test this hypothesis, eight students were randomly selected and asked to state how many hours they slept, on average, every night. Students then reported how many ounces of alcohol they drank on Saturday night. What statistical test would you use to determine if there is a relationship between the number of hours slept and the number of ounces students drank on Saturdays.
Pearson's r (ratio)
In an SAT prep book, you read a test-taking tip that states, "Individuals who read more books for fun will score higher on the verbal section of the SAT that those who do not." To determine whether this is true, you call high school seniors, randomly selected from a school directory, and ask them how many books they read per month for fun during their junior year, and what they scored on the verbal section of the SAT.
Person's r - relationship between books read and score on verbal section of SAT
Follow Up Tests for One-Way ANOVA
Planned comparisons vs. Post Hoc Tests
File Drawer Effect
Published research is a biased selection of all research
Total Variance
Systematic Variance (between groups) + Nonsystematic Variance (within groups)
Cohen's d (t-test)
To calculate you need: -mean -SD -sample size *you need all three for each group
The Between Subjects Factorial ANOVA
Total Variance --> Between Group or Within Group Between Group ---> Main Effect Factor A, Main Effect Factor B, A x B Interaction *Between subjects Factorial ANOVA w/ two IVS
A recent study suggested that females are more likely to eat chocolate than males. Each participant was presented with a Hersey's Kiss or Starburst. The male and female participants were then asked to choose which candy they would rather eat. What statistical test would you use to analyze your data?
Two-way chi-square
One-Way Between Subjects ANOVA
Used for a *SINGLE* IV with 3 or more different groups (between group stats) -*between groups (multiple groups, but with ONE IV)* -The one-way ANOVA formula will generate a single F score, and *larger* F scores are indicative of statistically significant group differences
One-Way Within-subjects (Repeated Measures) ANOVA
Used for a *SINGLE* IV with one group of participants who complete three or more conditions (within group stat) -*WITHIN GROUPS (ONE IV, WITH THREE OR MORE CONDITIONS)* *total variance = systematic variance (btw groups) + systematic variance (btw Ss) + nonsystematic variance (errors)
In social psychology, hypocrisy induction is used to try and change people's behaviors. The idea is that if you make a person feel like a hypocrite by having them think about his or her inconsistencies, he or she will want to change the behavior to correct the disparity and feelings of cognitive dissonance. In a recent study, researchers attempted to increase recycling behavior through this method. They had 200 students create video clips in which they preached that "everyone should recycle 100% of the time." Half of the students were then given a hypocrisy survey where they had to list all of the things they failed to recycle in the past two weeks. The other half was given nothing. Two weeks later, everyone received a survey asking about recycling habits over the past two weeks. One of the tests was to see if those who took the survey recycled a higher percentage of paper (10%, 20%, 30%...100%) than those who did not take the survey. What statistic would the researchers use to see if there was a difference in paper recycling.
independent samples t-test
Researchers have proposed that students learn new vocabulary words better when they are required to use the new words in a writing sample, as opposed to simply memorizing them. Students were given a list of 50 new words, and they were asked to study the words for 2 hours. Thirty students were then matched to different conditions based on their ACT verbal score. One group of 15 students were given the list and asked to sit in a classroom to memorize the words. A second group of 15 students were given the same list of words, but asked to write a story using as many of the new words as possible during the 2-hour time limit. After 2 hours, students were individual asked to vocally recite as many of the new words as they could. The researchers predicted that the students who use the new words in the story would correctly recall more words than the students who simply memorized the words. Determine if there were any statistically significant differences between the groups.
paired samples t-test (because they're matched)
You hypothesize that eating mint chocolate chip ice cream makes people smarter. You give participants an IQ test, then give them mint chocolate chip ice cream, and then give them the IQ test again. Which statistic would you use to determine if the ice cream helped to improve people's scores?
paired samples t-test (interval: IQ so you can't score)
CI for Mean Difference
range of possible mean difference does not include 0 -put lower and upper tails of CI