Lecture Notes, Lecture Exam 3 PDF

Preview

Summary

Download Lecture Notes, Lecture Exam 3 PDF

Description

4/19/2016  Next Tuesday (4/26)  Ethics and review  Next Thursday (4/28)  Review (last meeting day)



Concept list:  Page 3, strike out these from the course entirely:  73 (survey research and design)  74  75  76  104 (margin of error)  Know these for exam 3:  Convenient sample procedures  Probability sampling procedures 



Not on third exam, but on final (will be covered 4/26)  84  85  86  98 (basic knowledge already)

Delayed control group design  Data is collected when?  Delayed control group designs (experimental and control group NOT done at the same time like normal ones)  Sequential testing of groups  Ex: IV happens (event happens, like earthquake, that defines experimental group) Experimental group  data collected from experimental (ex: helping behavior = DV)  Control group (find similar group)  data from control group (DV)  This would be like when a natural disaster happens; community under stress come together or competing for resources 

Specific Threats to Internal Validity Faced by Longitudinal and Cross-Sectional Designs 1. Selective survival This is intrinsic to both cross-sectional and longitudinal designs. This threat is more critical with older adult samples. This threat is associated with changes in the population composition across time because the weaker, less competent, and less adjusted individuals have typically died off. This makes it difficult to make any retrospective or prospective inferences because the population is NOT the same (at different times). Ex: lead pipes stun your IQ growth; older people were alive when they were used 2. Selective dropout This applies to longitudinal research only. This is the situation in which participants drop out of the study sample. They might, for instance, move away, lose interest in the study, die, etc. So individuals who continue to participate may be inherently "different". 3. Practice effects or retest effects This applies to repeated measures longitudinal designs where the same individual is tested and retested on the same psychological behavior and tested over a long period of time. The problem is one of participants becoming task- or testwise. Also, if the particular task or test requires the use of particular skills, then with practice gained from repeated testing over a long period of time, participants become very skilled. A vivid example of this is the Berkeley Growth Study. This was a longitudinal study on intelligence in the 1930's. Over less than 20 years participants were tested on the same or different versions of the same test more than 40 times. It seems highly likely that performance on these IQ tests may have been inflated by practice.

4. History, cohort, or generation effects This is a threat associated with cross-sectional designs. Cohort—is some group that has some characteristic(s) in common; usually thought of in terms of different age groups. Cohort effect—the variable by which the cohort is grouped confounds the IV.

What should be the research primary concern when non-equivalent control group design? Rival hypothesis; alternative explanations Advantage of multiple TS over interrupted TS is? Eliminates the history effect as an alternative hypothesis Primary distinction btw true and quasi experiment RA p. Dr. Krall is using a _____ They designated the two groups so non experimental, therefore, it’s a non equivalent control group Which can be experimental or non exp design? observational What is post facto design most like? Postdictive, correlational design Most susceptible to history effects? Cross sectional Cross sectional design are best as? Between subjects designs Longitudinal are more practical and feasible. FALSE Practicing effects are threat to which design? Longitudinal

Psyc Test 3 Non experimental research design  Quasi experimental  High in external, low in internal when compared to experimental  Preexisting groups  Self-selection  RA is difference btw true and quasi designs  (Nonequivalent naturally occurring groups) for allocation to groups  Non assignment into groups  Comparing two groups  Can hardly ever make causal inferences  Ex: smoking causes cancer; took 30 years bc all of the groups were self-selected to so the results were weak  Non equal control group  Multiple group designs GD*  Nonrandom assignment into groups  Primary issues is interpretation of results with non-equal groups  Enhancing interpretation  Matching  Moderator variables  Pretesting  Statistical control  Examples  Delayed control group designs (exper and control group NOT done at the same time like normal ones)  Sequential testing of groups  Ex: IV happens (event happens, like earthquake, that defines experimental group) Experimental group  data collected from experimental (ex: helping behavior = DV)  Control group (find similar group)  data from control group (DV)  This would be like when a natural disaster happens; community under stress come together or competing for resources  These are TWO diff groups  Mixed factorial designs  One or both groups must be preexisting  We were not selected into groups-we were conveniently sampled  Ex: trait and state  Trait variable has to be the between bc you are either anxious or not (tall short) (white black)  This is what makes it a Q design; you can’t assign trait anxiety  State conditions can be repeated though, so state var is the within  CONTROL TECHNIQUES ARE NOT PARTICULAR TO EXPERIEMENTAL DESIGNS  GOT IT WRONG ON TEST 2  Designs without control group Aka single group designs (the control in this design are the scores before the intervention/event)  One group.  Interrupted time series design  Using a single group to assess an intervention  Collect pre intervention points, monitor presences of intervention, then collect post treatment data points  Ex: attitudes, then incident, then attitudes again to see if event interrupted attitude trend line  Aka trend analysis; looks like line graph  On notes; two interventions on the unemployment rate increases slightly, to 5.0 percent



Ex: if you have a line graph, that is horizontal then spikes after event. Enter control group that doesn’t experience event that caused the spike, and they are the control group now  Researchers don’t RA and the groups are naturally occurring  Disadvantages:  Always susceptible to history effect  Also, can be attributed to a basic employment cycle overtime  Can control; if you use a control group, you now have a multiple time series design  But it is no longer a design without a control group = added to solve the history threat

 Q designs  Non-equivalent control groups  Delayed control group design  Mixed factorial design  Designs without control groups  Interrupted time series design  Repeated treatment design  Confounding variable always result in uninterpretable designs  Pre-existing differences will always results in uninterpretable designs  Which has larger preexisting differences?  The chart that has a bigger difference on the pre-test side (looks like “7”), therefore, has less interpretable scores 

        

Topic 6  11 questions Topic 7  6 questions Topic 8  7 questions Topic 10  9 questions Topic 11  8 questions NO TOPIC 9, TOPIC 12 after the exam (4/26) ITEM EVERYONE GOT WRONG:  Control is unique or peculiar to experimental designs. Ans: FALSE  RA is unique or peculiar to experimental designs. Ans: TRUE

 4/12  TOPIC 10  Longitudinal and cross sectional designs (  these are more correlational, they are NOT experimental bc groups are naturally occurring  These two allow us to look at relationship between two as function of time  Used in developmental and gerontological (sp?)  Age and cognitive abilities  Age and self esteem  Age and contentiousness  Age and agreeableness, and emotional stability  Cross sectional  “quick and dirty” studies bc they can be done quickly bc there isn’t any re-administration  Disadvantage; history threats (each cohort has a different history)  Also, selective drop out and survival  Ex: unhappy people drop out (die) sooner, so it looks like happiness increases with age  Only way to get rid of it is to switch to a longitudinal design  Advantage; no practicing effects bc different groups  Goes across generations (cohorts) to look at interest  Ex: in 1980; 10, 20, and 30 year olds are put in 1950, 1960, and 1970 cohorts.  Between subjects  Several people are tested during one year  Longitudinal  Ex: stays with same cohort (1970) for thirty years (1980-2000)  More resource demanding than cross sectional  Within subjects; characterized by fact people are tested on multiple occasions  Advantage; no history threats bc same cohort  Disadvantage; testing and practicing effects bc of repeated administration  To try to get rid of these:  1) try to use alternative forms of test (problem; can’t attribute score change to only age)  2) switch from within (same person) to between (other people) design  To do this, you chose different people within the same cohort / HAS TO BE RANDOM THO  Disadvantage; drop out (deaths) aka attrition (mortality; threat to internal validity); exps are ran by labs and not one researcher bc of time  To fix these;  1) switch to between design  If they study is within subjects, and you follow the same cohort over time, the IV (scores from tests) will be the same on everything except the variable age  Bc of this, everything except age is held constant. Therefore, some people say inferences are plausible  (but answer is NO; NO CAUSAL INFERENCES)  Both cross and longitudinal are susceptible to selective survival (older adults; changes in the population across time bc weaker, less competent, les adjusted so they die off)  Ex: lead pipes stun your IQ growth; older people were alive when they were used  Comparing cross and longitudinal  Correlational designs; predictive, concurrent, and postdictive (this one sucks)  Longitudinal is predictive, cross sectional is concurrent  Predictive is superior design, so longitudinal is better too  Bc these are experimental, we cannot make causal inferences using either of these studies

 Correlational designs, are high in external validity, but low in internal validity  (just measure variables and look at relationships between the two)  (allows these to be used more in field then in lab where we have to mess with stuff)  Time lag  Hold age constant and looks at differences between cohorts  We won’t need to use these, we just need to know what they are

 4/12  TOPIC 11  Meta-analysis (MA) 1. In terms of impacts on field of science, and frequency of use, it is most popular 2. Secondary research method (vs. primary research methods (i.e., everything we have discussed to date)) a. Source of data/data points are results of other primary studies 3. Set of statistical methods used for quantitatively aggregating the results of several primary studies to

arrive at an overall summary statement or conclusion about the relationship between specified variables.  Advantage;  More standardized and relatively more objective  Magnitude of effect is important bc  Ex: if trying to meta-analyze several primary studies  You have to be able to take results and convert them into a common metric  if you’re gonna add lots of shit together, you have to get them to be the same thing aka effect size metrics aka r and cohens d  So it takes all those studies and makes those results r and d’s  To characterize stat significance of research, we look at effect size metric instead of statistical significance (p)  If the primary study looked at diff between means? D is used to meta-analyze results  If the PS looked at the associations of variable? R is used to meta-analyze results  Sooooo with that; not all primary studies are equal  Ex: studies with large sample sizes are BETTER than those with tiny ones  Therefore, we wanna give these more weight than the small ones  MA sample-weights the results by sample sizes, and then gives those with large ones more weight  Therefore, all the sample size differences between primary studies is controlled for  Math; find all sample sizes (n) find all effect sizes (r or d) multiply (n and r or d) then divide that by N  But hey, we don’t need to know how to do it  Disadvantage;  1. MA can only be as good as the primary studies its looking at  2. Even within primary studies, there are differences between studies. So MA is therefore comparing apple and oranges  3. Limited number of primary studies available;  We are trying to use ALL of the PS, NOT a SAMPLE of them  4. Selection of primary studies for inclusion  Have to have a criteria for selecting the studies  5. File drawer problem;  If you rely on only published studies, you aren’t including studies that weren’t. therefore, bias set of studies  6. Judgement calls;

 TOPIC 8  Predictive correlational designs are the strongest  They allow us to make strongest relationship between the antecedent and the consequence because of the appearance of temporal precedence (which one comes first)  Postdictive are the weakest  Selective...