1 / 24

Experimental Research

So far. Can examine people as they areCan look for between-group differences or within-group differencesExamples of questions can address?Key question cannot address what words not allowed to use?. Why not?. If just take people as they are and measure them, and there are differences, why are there differences?Ex: if we use a static groups design and find that evening students do better than day students why might that difference be there?.

lael
Download Presentation

Experimental Research

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


    1. Experimental Research Corey and Lachter chapter 4

    2. So far… Can examine people as they are Can look for between-group differences or within-group differences Examples of questions can address? Key question cannot address – what words not allowed to use?

    3. Why not? If just take people as they are and measure them, and there are differences, why are there differences? Ex: if we use a static groups design and find that evening students do better than day students – why might that difference be there?

    4. What to do? We want to make sure that people are as equal as possible on all other variables that could be causing a difference This way, if the only overall difference between groups is which group they’re in, we know that this must be the cause of a difference on an outcome variable

    5. First, some terminology Always have at least two variables in a hypothesis May refer to as predictor and criterion variable in non-experimental study In experimental study, predictor = independent variable and criterion = dependent variable

    6. The terminology Independent = cause will be testing Dependent = effect will be testing Study includes both experiment and non-experiment “Experiment” should only be used if a variable is manipulated

    7. Back to making the groups equal How can we make sure that groups are equal on everything but independent variable? Measure every variable in the world? Key = not ensure that every person is equal, but that, overall, the means of the groups are equal

    8. How to do this? Key = assign people to groups How? Random assignment How it works Not to be confused with random selection

    9. Experimental design and research question and conclusions When is an experiment needed? News, conclusions, and studies Frequently, causal language is catchier News frequently presents results of non-experimental studies as implying causality Prison recidivism and marital status

    10. Other considerations in experiments Between versus within distinction still applies Want to have half participants have one order first and half have the other order first Can look at interaction between order and level of variable

    11. Other considerations in experiments Interaction effects still apply SE and valid/invalid test Amount of exercise and mood, with how enjoyable exercise is

    12. The limits of experimental research What sorts of variables cannot, practically or ethically, be manipulated?

    13. Another advantage of experimental designs Help minimize confounds (when one variable changes systematically with the independent variable, or predictor variable) When this happens, the study lacks internal validity (internal validity = study is testing the hypothesis that you think it is)

    14. Formalizing control issues Confounds are such a threat to internal validity that a list of confounds to watch for has been developed Here is the list, will talk more about each: Selection bias History Maturation Effects of testing Statistical regression Experimental mortality Instrumentation Subject bias Experimenter bias

    15. Selection bias One type of participants end up in one group, while another type end up in another Random assignment is designed to help minimize this

    16. History Factors outside of the laboratory setting that may affect the dv For instance, if a within-subject design is used to examine the effect of student discussion groups on attitudes toward school violence, and there is a school shooting after the non-discussion group measure and before the discussion group measure Randomly assigning participants to different orders of levels of the iv in within-subjects designs helps guard against this confound

    17. Maturation Changes in participants due simply to the effects of time Like history, especially a problem with within-subject designs in which there is not random assignment to different orders of presentation of the iv

    18. Effects of testing Again, this is especially a problem in within-subjects designs This is another problem with people being people Simply measuring the dv once can affect the result you’ll get the next time you give the measure People may get better at a test If people are given the same test more than once, they may assume the experimenter expects a different answer – why else ask the same question twice?

    19. Statistical regression Regression to the mean Each person has an individual mean on various measures – e.g., individual mean anxiety level over time If you catch someone when they happen to be at a high point, they may give a lower response on the next measurement that’s simply due to them going back to their mean level, rather than due to any manipulation in the iv Example: extremely tall parents tend not to have children who are taller than they are

    20. Experimental morbidity (or experimental mortality) Some conditions of an experiment may simply be more fun to be in Participants have no obligation to participate in your experiment, and they can drop out any time If more participants drop out of one condition, or if more of one type of participants drop out of one condition, this raises a problem of internal validity Are your results due to your iv, or are they due to the fact that you only have data from who did not drop out?

    21. Instrumentation Change in how well the instruments you use to measure your variables work over time Especially a problem if level 1 of iv measured first and level 2 measured second, since the instrument worked better for one level than for the other level Could also be factors such as RAs running the study getting tired

    22. Addressing these concerns Random assignment to condition, or random assignment to order of conditions can help correct these possible confounds Experimental mortality is harder to correct, short of preventing participants from dropping out of study

    23. Two more threats to internal validity Subject bias: participants bring expectations with them to the experiment. These expectations, rather than the iv, can affect their behavior e.g., one group given what think is alcohol – expectations can alter behavior (study with making people think getting drunk) Same as placebo effect Problem if expectations of one group are different from expectations of another group Not a problem if expectations can play the same role in both groups (single blind – participants don’t know what level of iv group they’re in)

    24. Two more threats to internal validity Experimenter bias: if experimenters are aware of the hypothesis, and of which condition participants are in, they may inadvertently treat participants differently, or measure the dv differently e.g., Morton and BB’s To correct: double blind: neither subjects nor experiments know what group each subject is in

    25. Bottom line The more control the experimenter has to ensure that the only systematic difference between groups is the level of the iv, the higher the internal validity, and the more confidence the researcher can have that any effects found are due to the iv, rather than to some other factor Minimizing potential confounding variables helps isolate the iv from other possible causes of differences in the dv

More Related