This preview has intentionally blurred parts. Sign up to view the full document

View Full Document

Unformatted Document Excerpt

DO HW: THAT SILLY HYPOTHESIS SHIT....write a simple hypothesis...she'll show us how to do a comparison using excel later...don't worry about learning how to do it on your own...DUE THURSDAY 10/12/06 10/10/2006 The scientific method Theory identify general rules of how things in the world works o explains why something happens o organize o predict Broad Relevant - Theory Hypothesis Study Statistical Analysis Replicate Hypothesis o Testable prediction Specifics need to be specific to test things Falsifiable has to be open to be proven wrong Null hypothesis this kind of hypothesis is BAD o Predicting no relationships o Cannot be proven o Only reject Study o Test of hypothesis Naturalistic observation Hawthorne Effect...the very presence of observation will chance the dynamic of how the subject interacts....usually high external validity Laboratory low validity environment, but can manipulate the environment Questionnaire Statistical Analysis o Quantitative data o Interpreting results Is hypothesis supported? - - - Correlations - How are two variables related? - Correlation o statistical measure of relatedness Positive correlation variables increase or decrease together / - up slope Negative correlation one variables go up, the others goes down. \ -down slope Correlation coefficient - numeric measure of how well points fit line - rage -1 to 1 Coefficient of determination - strength of correlation Correlation DOES NOT imply causation - direction of relationships unknown Experimental studies - experiment o manipulation effect - Bandura's social learning experiment - Independent variable o What gets manipulated - dependent variable o what gets measured Independent variable dependent variable Experimental group Control Group Random assignment Read summaries for each chapter 1-5 Also do those silly labs, 10 questions based on and another 40 questions based on the textbook. Bring a scantron UCD 2000 Don't read over the notes, look at the stupid summaries or else you'll never get it right because she doesn't use easy multiple choice. Measuring Behavior Reliability and Validity Creating operational Definitions What to measure - verbal behavior - overt action planned or purposeful expression - nonverbal behavior - physiology Measuring behavior - create environment - participants perform behavior - record behavior - qualitative (qualities of something) to quantitative - measurement error any inaccuracies in measurement Measurement Error Random - inconsistent variations - not due to construct - unrelated to hypotheses - random results due to chance Bias - systematic variations (find things that aren't really there) - systematic - threaten validity Reliability - stable, consistent scores - free of random error Validity relies on reliability Environment Errors - Standardization o Standardization versus humanism Participant Errors - Random error o Large sample help stop the fluctuations - participant bias o demand characteristics unintentionally influences participants unobtrusive measurement use distracter measures measure physiology create costs or benefits o social desirability behaving to look good demand characteristic techniques anonymity social desirability measure Reliability - test-retest reliability o measuring a behavior twice o assume behavior to be constant test retest coefficient - observer o multiple observers interobserver agreement (% of times researches agree) interobserver reliability (correlate observer ratings) Internal Consistency Choosing the best Measure and finding the best subjects Sensitivity - Detect differences in participants - Avoid behaviors that are resistant to change - Avoid all-or-nothing measures - "how much" instead of "whether" (yes or no) - add scale points to rating scales o 5-7 point scales - pilot test measure o ceiling effect (all people rate something really high, leads to no differences and probably caused by socially desirable effect) o floor effect ( give more options to fix his problem to make it more sensitive) Likert Scale - please rate the extent to which you agree with the following statements using the scale below: disagree strongly 1 disagree moderately 2 disagree slightly 3 Neutral 4 agree slightly 5 Agree moderately 6 agree strongly 7 - Using the scale below, please how true the following statements are about you: Not true at all Slightly true Moderately true Extremely true 1 2 3 4 This is a demand characteristic, forces people to make a decision that isn't Variance - differences in participants - essential for statistical analysis - no variances = no correlation Types of Questions - Do two groups differ? nominal - Does one group have more? ordinal - How much more? interval - Three times a much? ratio - o What is the functional relationship (exact measure)? quantitatively different questions Scales of Measurements - nominal o numbers represent groups o identify categories - ordinal o meaningfully order o know which is greater o do not know how much greater - intervals o numeric intervals represent equal distance - ratio o numeric intervals represent equal distance o absolute zero o relative magnitude Who to Measure? - Population o Group of interest - sample o who is measured o representative Sampling Strategies - convenience sampling grab whoever you can - quota sampling - random sampling o random sampling error - stratified random error - nonresponse bias 10/26/06 Statistics Purpose - describe sample o not population - sample characteristics - external validity - abnormalities o outliers Central Tendency - mean o common average o sum of scores/number of scores - mode o most common score - median o middlemost score Variability - range o differences between highest and lowest - interquartile range o percentile o difference between the 75th and 25th percentile Variability - average squared deviations from the mean o find out how far each value differs from the mean o then the average from differences (squared), then get the average of that (Sum of squares) o Variances = Sum of Squares/N Mean Mean deviations Mean deviations squared Sum of sum of squares/sample size = variance - Standard deviations = Variance Normal distribution - frequency distribution which occurs often in nature o random sample of entire population o a bell curve o 1st standard deviation is about 68% o 2nd standard deviation is 95% o 3rd standard deviation is 99% Frequency Distribution -usually positively skewed distribution | Graph Representations - bar graph o categories of responses o frequencies within categories - line graph o changes in score over time - stem-and-leaf chart o represents actual data points - scatter plots o able to see relationships a bit better o find out a correlation Two-group design 1. needs to be identical group a. selection bias i. Self assignment may have diff motives. Go 2 study session b/c motivated, while other is failing ii. research assignment research picks assignment iii. arbitrary rule assignment- half goes here others half there iv. Matching match kids in exam rooms. Person 1 goes 90 in 1 group and person 2 get 90 in group 2. Other level may differ IQ motive v. pretest matching b4 decide who goes in candidates test 4 vi. 1 selection by maturation interaction not every 1 strength changes the same way. Ex test ever 1 in grade 2 then test again in high school. But what if 1 group is boy and other is girl Pretest Post test Design - assumes treatments is only cause of change o participants change independently nothing 2 do with variable Reasons participants change - maturation expected changes as one grows up - environment things in the environment that can change us - testing effect can change a person's own view on how thing work Measurement Error - instrumentation o any bias resulting from different instruments - Attrition problem of participants leaving. Sometimes you just don't know why participants are leaving. Can really screw up your measure - regression effect (extreme scores that change to the mean) o regression towards he mean Extraneous Variables - measure - statically assess influence The Simple Experiment Manipulation effect Independent variable dependent variable - Experimental group o Manipulation - control group o comparison group - larger effects are easier to detect o Example: looks at the effect of pornography contains scenes that have rape themes. Shows that those themes decrease sensitivity to rape scenes. To what effect does that have on a person. One group watches those themes, oppose to non-themed - assigning participants o random assignment o participants remain independent (so each participant won't be influencing each other) o lessen random error with more participants - Standardization o Measures o Participants (use homogenous participants increase internal validity but lowers external validity) Analyzing experiment Data - determining effect of manipulation o compare manipulation and control groups o comparing dependent measures - descriptive statistics o what is the difference? (this is what happened to one group compared to this other group) - inferential statistics o is difference meaningful? (whether or not the differences are significant) Statistical significance - likelihood that results represent reality - p-value o probability of results being due to chance o 100% confident 0% due to chance p - .00 o 95% confident 5% due to chance p < .05 - different from effect size o how large that difference is - can be theoretically insignificant o doesn't mean its significant because the study was conducted terribly - may refute hypothesis Nonsignificant Results - failed to reject null hypothesis - cannot prove null hypothesis o cannot prove no effect - common errors in interpretation o nonsignificant means no effect o there was an effect despite nonsignificant (can't say it) Errors in Statistical Significance - Type I Error o rejecting null when treatment has no effect o "crying wolf" - Type II Error o fail to reject null when treatment has effect o "failing to announce the wolf" How to fix these problems - Type I o Reduce by lowering p-value - Type II o Reducing by increasing power (we're trying to make the finding or study stronger) Reduce random error by increasing sample size Creating lager effect Analyzing the simple experiment - calculate mean for the each group - compare group means t-Test - is the difference larger than would be expected by chance o t-value difference between group means within group variation - degrees of freedom o number of participants minus number of groups The bigger the t-value and the bigger the degrees of freedom the better for your results Group Mean Differences - difference at group not individual level - outlier influence o can make you find things that are not there o report deletion of outliers Frequency distribution Negatively skewed distribution where the mean is to the left of the mode Positively Skewed frequency distribution is where the mean is to the right of the mode 11/7/06 Multiple-Group Experiment - compared more than two groups - reduce number of experiments - reduce number of participants U-Shaped Relationships Performance Low Med motivation High U-shape can explain a lot of conflicting data (Need at least 3 variables) Functional Relationship - shape of relationships linear Dependent Variable Non linear Independent variable Nonlinear Relationships Quadratic (U-shaped) Cubic (double U-shape AKA sine) S-shape (plateau incline) J-shape Construct Validity - confounding variables o multiple manipulation groups o multiple control groups empty control placebo group group - hypothesis-guessing o people are acting on what they think the hypothesis is Advantages - reduce number of experiments - reduce number of participants - detect functional relationship - increase construct validity EXCEL SHIT T-test for comparing two groups Correlation click on a blank cell for correlation This is r value =pearson(array1, array2) gives correlation coefficient Look at the sheet ...compute t-value....use the chart in the book ~400....degrees of freedom closest to number but not over 11/9/06 EXAM REVIEW: 11/9/06 @ 6:10-7PM 2 WELLMAN ALSO NEED TO KNOW THE "SETUP" TESTS MULTIPLE GROUP STUFF Survey Research Using self-report measures Using Self-Report - disadvantages o low internal validity o response inaccuracy Response Inaccuracy - don't know the answer - don't remember the past o retrospective reports (people forget the information) - cannot predict the future - don't choose to be accurate o social desirability o response set (don't read the question and give just the same answer) Advantages - access to thoughts and feelings - inexpensive - flexible Survey Instruments - questionnaire - interview Questionnaires Technique Self-administer Investigator-administer Advantage - inexpensive - easy - anonymity - clarification Disadvantages - low return rate - biased sample - confusion - less perceived anonymity Psychological Test - established questionnaire measure of construct - pre-testing (measures quality of test) - reliability - validity - objective scoring - counterbalanced (opposite question to another) - standardized Interview - advantages o direct communication - disadvantage o time consuming administering training - structured o fixed-alternative questions - semi-structured - unstructured lots of problems and bias Technique Face2face Telephone Advantage Detailed information Intimate Personal More anonymity Convenient Faster Group dynamics Cheap Huge sample size Do it whenever Disadvantages Less anonymity Costly No visual cues Focus group Online Biased data Total biased Only computer users Less certainty Writing Survey Questions - avoid leading questions - word question to avoid social desirability - order questions carefully - avoid double barrel question (2 questions within a question) - avoid long questions - avoid negation (do you disagree with not support....bad) avoid irrelevant questions avoid big words avoid slang INFORMATION FOR THE NEXT EXAM MULTPLE GROUP ANALYSIS ANOVAs Comparing Multiple Groups - multiple t-test? o Maximizing chance o Can't run multiple t-test because 5/100 will be do to chance alone - variability between groups Comparing Multiple Group Means - between group variance o random error o treatment variance variance in how you treat the group - within-group variance o error variance Analysis of Variance - ANOVA o Comparing three or more groups Between-group variance = F ration Within-group variance F ration = random error + possible treatment effect random error - Significant F ratio o Value must be greater than one o Degrees of freedom Within-group Number of participants minus number of groups between groups number of groups minus one Interpreting Significance - group variance not due to chance o how the groups differ from each other? Post hoc test (testing something after the fact) Run test after F is significance - what is the functional relationship? o Post hoc trend analysis - how large is the effect o estimate with group mean differences o eta squared actual difference/potential difference (ratio tells you how much variance are in your groups from being potential) 11/30/06 Other Experimental Designs Matched-Pair and Within-Subject Designs Matched-Pairs Design - requires few participants Produce - matching variable - related variable - matched pairs - randomly assign pair member to group Considerations - finding a matching variable o similar measure o correlated measure - power o ability to find group differences o matched-pairs increase power o good matching is essential - external validity is lost o because we tried to get rid of individual differences o increase this by...using good matching quality can have diverse samples if you have a good matching quality o decrease selective attrition (people leave, but not randomly) testing effect (getting exposed to a measure beforehand can change people) - construct validity o pretest suggests your hypothesis at all, it can cause a problem Data Analysis - simple experiment o t-test independent t-test assumes group independence - matched-pairs design o dependent t-test accounts for the group relatedness Pros & Cons - advantages o more power o less restricted population - disadvantage o alert participants to hypothesis o Using deception in anyway is a bad idea o Cannot generalize to drop-out population Within-Subjects Design - repeated-measures design o every participant receives all levels of manipulation (one group gets all manipulations) Considerations - increase power with less participants o no fluctuations between participants o decreasing random error - order effects o variations in response behavior due to order of conditions o decrease internal validity Order Effects - practice effect o performance improves on later measures due to experience and learning minimize with practice prior to experiment - fatigue effect o performance gets worse because of fatigue or boredom minimize with shorter and more interesting experiments - treatment carry over effect o earlier treatment effect later treatment minimize by increasing time between treatments - sensitization effect o performance changes due to hypothesis guessing minimize by preventing participants from noticing treatment changes Minimizing Order Effect Overall - less conditions you have the better - change sequence of the conditions to balance out order effects o randomize Counterbalanced Within-Subjects - create one group for each order o two conditions yield two groups o randomly assign participants to groups - factorial design o condition what is the overall effect of condition A or B o Order what's the effect of the sequence AB or BA o Interaction Order or trial error Randomized Within-Subjects - matched pairs - each participant in pair gets different sequence - data analysis o dependent t-test o within-subjects analysis of variance Choosing the Right Design - one independent variable o matched-group design if it is easy to obtain matching variable when participants are scarce o within-subjects design if order effects are not a problem need powerful design actually interesting to what happens when a person if exposed to multiple treatments (external validity improved by exposure to both treatments) o counterbalance design (2x2) want to be able to detect the order and sequence effects have to have a lot of participants can't be a problem with exposing people to multiple treatments (can expose participants to both treatments) o between subjects design if order effects are a problem have enough participants external validity improved b exposure to only one treatment Quasi-Experimental & Single Design (you can't always get what you want) Inferring Causality - co-variation o relationship between changes in independent and dependent variables - temporal precedence o independent variable occurs in time before dependent variable - eliminating spuriousness o changes in dependent variable are only due to independent variable Single-n Design - case study o no random assignment o no inferential statistics - random variable affect all groups equally (yes we know people have individual differences but by using random assignments, we know that random groups distribute the variations) inferential statistics estimates effect of random variables covariation temporal precedence spuriousness o keep non-treatment factors constant A-B Design - most basic single-n design - establishing a stable measure of the dependent variable o stable baseline this is the "A" part of the design - post-treatment dependent measure o this is the "B" part Base Line Stability baseline Treatment Battling Spuriousness - eliminating random error o by keeping non-treatment variables constant - potential source of error o between subjects variability not an issue - within-subject variability o stable baseline o stable environment A-B Design Limitations - requires stable baseline o control factors contributing to random error o difficult to identify factors - Within-subject design effects o Practice effect give practice before measuring o Fatigue effect give breaks, make it interesting, so they're not bored out of their mind o Sensitization effect These are testing effects - maturation o natural biological or psychological changes A-B-A Design - reversal design o baseline (A) o treatment (B) o baseline (A) - treatment effect o removing treatment results in return to baseline A-B-A Design Limitations - cyclical maturation o A-B-A-B-A design - Carryover effects o Whenever the treatment's effects persist after removal Multiple-Baseline Design - Measure baseline for several key behaviors - Only behavior targeted by treatment should be affected Validity - internal validity o keep non-treatment variables constant - construct validity how well we think we measured what we wanted to measure o threatened by sensitization placebo baseline subtle differences between phases - external validity o design with homogenous populations o account for the individual o replication see if it valid for other people Quasi Experiment Design - similar to true experiments o don't use random assignment o cannot control non-treatment factors - account for spurious variables Stanley's Spurious Eight - maturation o natural biological or psychological changes - testing o previous exposure to dependent measure - history o events in outside world - instrumentation o changes in measure - regression o extreme scores move closer to the mean - attrition o selective participant drop-out - selection o groups are different before you get to the treatment - selection-maturation interaction o think your groups are the same when they start out but they mature different rates.....groups grow apart Quasi Experiment - control for spuriousness when possible o instrumentation - measure potential spurious variables o regression analysis compares relative influence of different variables Factorial Design - examines combined effects of two or more factors - 2x2 factorial design o two factors (IV) o two levels of each factor (IV) 2x2 factorial design Two independent variables No exercise 2 levels of exercise Exercise No diet No exercise Diet unique effect of each factor o main effect combined effect of both factors o interaction No diet Exercise Diet Main Effects Simples main effect - difference between groups - group mean differences - Overall main effects o Average of factors simple main effect overall main effects o simplicity o greater confidence that effect generalizes Interaction - combined effects of both factors - effect of one factor depends on the level of the other factors Moderators - influence effect of another variable o intensify o weaken o reverse - personal characteristics - context Potential outcomes (2x2) - one main effect, no interaction - two main effects, no interaction - one main effect, interaction - two main effects, interaction - no main effect, interaction - no main effect, no interaction One main effect and no interaction How do gender and diet affect weight loss Main effect - overall main effect o diet Interaction - compares simple main effects o none ... View Full Document

End of Preview

Sign up now to access the rest of the document