Class Notes (836,289)
Canada (509,725)
Psychology (3,518)
PSY100H1 (1,627)
Lecture

PSYB01 - Lec 8 - Additional Research Designs (near-verbatim)

9 Pages
110 Views
Unlock Document

Department
Psychology
Course
PSY100H1
Professor
Connie Boudens
Semester
Fall

Description
Lec 8 (Nov 15, 2012) Additional research designs Quasi-Experiments (aka field experiment) – type of experiment but doesn’t meet all requirements for a really tightly controlled experiment; some differences but they have nature of experiment (which is a true experiment) Done in naturalistic setting • Less control of extraneous variables. • Variety of reasons, usually because you don’t have random assignment; most of the time, this is what’s missing • But other things may be that you don’t have control group • May be the case that you have a confounding variable but you’re not able to separate the effects of the compound Some common features of quasi-experiments: • matching instead of randomization • sometimes can’t do randomization; match participants in control and experimental group on whatever variables are key to your study • time series analysis • type of quasi-experiments • unit of analysis not people • can be organization, group, program or something in this nature One Group Pre-post Test  Say you wanted to look @ training of staff in school cut down instances of bullying o Would start by counting the instances of bullying over certain period of time; do staff training and then after see if instances of bullying gone down  Appropriate for research where you wouldn’t be able to get two groups of participants that you would randomize o Since program will affect entire school, you won’t have a group to compare the results to; so you can just assess instances of bullying before and instances of bullying after Threats to Internal Validity If you have a really well designed study with control group and control for extraneous variables and random assignment – then should have strong internal validity (meaning that you have very strong evidence to tell you that the reason for the change has to do something with the manipulation of the independent variable). But with quasi-experiment design, will have some problems making that conclusion  History o Anything that happens btn pre-test and post-test that could potentially affect the outcomes o Ex) doing above study in a certain school and then Amanda Todd event happened during that time – would find that it would have some effect on bullying  Decreased instances of bullying wouldn’t be because of staff training but rather because of the incident that happened o It’s not past history; not from BEFORE your experiment; it is what happens BETWEEN your pre- test and post-test  Maturation o Change in the subjects over time; usually some sort of natural change o Ex) Gr 8 and gr 9 – look @ instances of bullying – may see changes due to just the factors that they’ve gotten older o Anything that you’re looking at over period of time; not just over long period of time; can be much shorter period of time  Testing / Repeated testing o Repeated measures designs; participants are exposed to all diff levels of independent variable o Can get bored, practice/learning effects o Testing itself can have an impact; ex) what are the instances of bullying in school – ask kids to fill out questionnaire; just filling out the questionnaire will have more of an impact – will make ppl more aware  Mortality o Ppl dropping out of the study (not just b/c they died, they left the study for some reason) o The problem is not just the ppl who drop out, but if the ppl who drop out are a certain type of ppl  Ex) smoking cessation: hypnosis as treatment – smokers are participants, there are some ppl who will go back to smoking; wouldn’t be a problem if they were randomly selected but they are actually going to be specific individuals – might be ppl who are less suggestible – stable personality trait that can actually be measured; ppl who drop out of study are ppl low on suggestibility – means that your results aren’t going to be as valid b/c you haven’t tested on variety of ppl  It was successful on ppl with high suggestibility – but won’t know that unless you tested for that in advance o It’s the potential that SPECIFIC TYPES of ppl will drop out of your study  Regression to the mean o Happens when you take baseline/pre-test measure and whoever you’re measuring scores really high or really low for whatever reason o Over time, it regresses back to mean for the group o Ex) schools in particular area have certain # of avg bullying instances in a year – with the study you want to find a school that is above the mean to do your intervention – to figure out if the intervention actually works o You see that the average instances actually decreases to the mean; but it’s not due to the intervention, new kids get used to each other, so less bullying, the bullying instances regresses back to the mean o It looks like the intervention was successful for that school but it was just that the instances was very high for that school and it regressed back to the mean o Illustration:  In pre-test measure, the score for the particular instance that you look @ = low  In post-test measure, there’s a change but not necessarily due to the intervention  Just returning back to the personal baseline Another type of experimental design: Non -equivalent control group design  Same thing as above but with a control group ; sometimes and optimally you would do this with random assignment o Randomly assign ppl to control and experimental group – so know that individual differences would even out between the groups but you can’t always do that o Another way to do it is to find a school that is similar in important characteristics (student age, size of school etc;) and this can be the control group o Compare the diff between the two schools – problem with this is that there’s selection bias = some pre-existing difference between groups; have to try best to ensure things are as equivalent as possible but since it isn’t random assignment, there may be some stuff in there that may affect the results  But try to match groups as closely as possible Time Series Analyses  Type of research design you can use when you can’t randomize the participants but you can get assessment of pre and post treatment or various points during the times  Graph = shows control time series o Here, they looked at greater traffic accidents over time; where there’s P, that’s when speed limit imposed o Traffic accidents = white line = in Connecticut; green line = control state o Controlled for weather conditions because similar regions and assume that there’s going to be similar traffic at similar times of the year o Control time series design because you have pseudo-control group o This is also interrupted time series design because it looks @ data at various time points; most basic time series design would only look @ data @ two separate points Single-Case Research Designs • Use only one case or group to investigate a specific phenomenon. • Not the same as a case study. • Because case study design is usually after the fact • Uses time-series design. • Take measurements of DV at various points in time • Take multiple pre and post-treatment measures. • With this type of study, you’ll introduce some sort of intervention or manipulate IV in some way Advantage of Small-N Design • Participants from hard to find populations (unusual illness, or someone who’s about to have unusual experience) • N= # of participants • N of 1 = only one participant • Look @ his changes in cognitive functioning before and after a certain event • Results easy to interpret (often no stats) • Can focus on helping one (few) participant(s) • Ex) phantom pain after amputation; how to help these ppl but this is pretty rare phenomenon; attention diversion intervention; took 6 measurements over time (3 baseline readings and 3 after receiving intervention) • Is really easy to interpret A-B-A Design  Take pills and then stop taking pills and then take pills again to ensure that it was the pills that worked o N of 1, ABA design o Ensure that there’s no residual effects during B stage when you take away the pills, it might be that you never go back to baseline even if you do take away the pills because you are more relaxed knowing that the pills are there when you need them o You don’t know what is the reason that you didn’t return to baseline; you can’t tease out the reason because you don’t have control group o With control group, can tell if it was because of treatment or if because of some other change Multiple Baseline Design • Testing a treatment effect when effect is irreversible. • If there’s a change that the treatment effect may be irreversible, then collect more baseline data – can be done in diff ways • Baseline data collected on: • 2 or more behaviours for same individual • Person who likes cocaine; measure tendency to take cocaine AND measure tendency of person to go to places where they had cocaine in the past • Same behaviour for 2 or more individuals • Same behaviour across 2 or more situations for the same individual. • ** for the first and third option, the person is basically serving as their own control Choosing a Research Method – all research methods have advantage and disadvantages *Experiments are given very prominent place in psychology because they allow investigations of psychological phenomenon to be very scientific; makes it much more like hard sciences, control variables and study relationship of cause and effect But all other research methods also have impt role in analyzing human/animal beh The choice is affected by: • Resources like time and money • Sometimes need answers more quickly, in a way that’s lot easier and cheaper • Ethical concerns • Also affects things like any research that involves harm (can’t randomly assign child abuse for ex) • The research question • Main reason that should affect your choice of research method • But that doesn’t always work out; sometimes you have to tweak question depending on what’s available Description: asking what or how many, getting ideas about why (preliminary ideas), developing theory – Descriptive research questions, then would use: (case study = 2 types or correlational design) • 1) Case study for unusual cases • Something unfortunate that occurred for ex • Ex) discover child that’s never been spoken to; happens rarely, but does happen – when found, then good for psychologists to figure out things; but can’t do experiment because this is serious ethical issues • Ex) Hostage taking situations – discovered Stockholm syndrome – hostage identifies with the hostage-taker • You’re not going to sit around for these things to take place, so most components of your study will involve using other research methods • 2) Case study for in-depth examination • Know something in advance • Look @ how process goes from beginning to end (ex: drug development from beginning to end) • Correlational study • Mostly addresses questions like ‘How is factor X related to factor Y?’ Prediction: if I know X, can I predict Y? • Correlational study Causal: Verifying the why or how • Experiment, quasi-experiment, field experiment • Can speculate a causal relationship based on data from other types of studies/research methods; but can’t know for sure unless you do experiment Each method has advantages and disadvantages; major factor that will determine which research method you’ll choose is whether you’re going for external or internal validity – major trade-off – you can’t really have both The best way that researchers ensure that they have both internal and external validity in
More Less

Related notes for PSY100H1

Log In


OR

Join OneClass

Access over 10 million pages of study
documents for 1.3 million courses.

Sign up

Join to view


OR

By registering, I agree to the Terms and Privacy Policies
Already have an account?
Just a few more details

So we can recommend you notes for your school.

Reset Password

Please enter below the email address you registered with and we will send you a link to reset your password.

Add your courses

Get notes from the top students in your class.


Submit