Psych 303
IV Groups that help establish covariance
Control - intended to represent no treatment also known as the placebo group when participants given treatment with no effect Treatment - the one given treatment
Within groups concurrent measures
participants exposed to all levels of an independent variable at the same time
Post/ pretest design
tests DV before and after exposure to IV
Post test design
tests the DV only after exposure to IV
Selection Effects
type of confound occurs when the kinds of participants of one level are systematically different than those of another level and results differ than they would if it werent the case
Expirimental IV
the manipulated variable. I looks like 1 (first in time) always on the x axis in a graph.
Experimental DV
the measured variable. how a participant acts on the measured variable depends on the dependent variable. always on the y axis.
Situation noise
the situation in which data is collected changes the variablility
Preventing Instrumentation threats
use post test only make sure pre and post tests are equivalent use manuals/standards for behaviors
Reasons for null effects
variables dont covary study not designed well enough
Noisey Data (error variance)
variance within the groups is less= smaller difference between groups therefore more similarity between groups, smaller the effect size variablity comes from measurement error, individual differences and situ
Temporal Precedence
when a variable is manipulated, temp precednce established because you know the causal variable lead to the effect v diff than associations
Insensitive measures
when it comes to dependent measures it if condmperable to hold fjb
Order Effect
when the order that the variables are given in effect their results eating apple and then chocolate(which is sweeter)
Covariance
Indicated by a difference in means between two groups levels of IV's can help establish a comparison requires manipulation of groups (3 types of IV's)
Individual Differences
Individual differences can effect results solutions are to test lots of people or change the design
Do all experiments need a control group?
No
Reasons for no within group difference
Noise: unsystematic variablility skews the data within groups
Causal claim terminology
causes influences, affect, makes, result in,
Solutions for measurement error
choose reliable (test retest, interrater and internal) forms of measurement measure more things
Types of within group design
concurrent measures repeated measures
Weak manipulations
how researcher operationalized the variables might be reason for null effects
Within-groups design
only one group of participants where each participant is presented with all levels of the IV
Testing threat
order effect because participants have taken test more than one time and therefore know what to expect
Demand Characteristics
Participants guess what the study's about and change behavior in expected direction
Placebo effects
Participants improve because of the drug/ treatment the beleive they are receivining
Instrumentation Threats to Internal Validity
- when a measuring instrument changes over time - when a coder for behavior changes standard of judgement - when different pre/post tests are used and they are inaccurate
Internal Validity
-ability to rule out alt explainatns (confounds) -must ensure that the IV manipulated was the one that caused change in DV
control variables
-any variable the experimenter holds constant on purpose -levels same for all participants -allow researchers to eliminate alt. explanaitions (internal validity)
Ceilings Floors and DVs
...
Ceilings floors and Iv's
...
Observer Bias
...
Within groups repeated measures
...
three rules of causal claims
1. Covariance 2. Temporal Precedence 3. Internal Validity
Design Confounds
2nd variable that varies systematically along with the iv and is an alt explanation for result
Solution to testing threat
Avoid pretest Use two different but similar tests Use comparison group
Threats to internal validity
Design confound, unsystematic variability, selection effects,
History
Expirimental group changes over time due to an external even that effects all or most of the group
Regression to the mean
Group whose scores are extreme at pretest due to random events will become more average at posttest
Which is better
Postest fine when random assignment guards agains selection effects. Also better when you dont want subjects to know what you are testing for or know subjects will just answer the same after Pretest good when you want to be extra sure that your IV is effecting DV
Experiment
Researchers manipulated (even if thats assigning people to either a control or experimental group) one variable and measured another
Reasons for no between groups difference
Weak manipulations Insensitive measures Ceiling and floor effects
Ceiling and floor effects
all scores of IV groups on DV are squeezed together either on the high end or the low end
Independent-groups design
different groups of participants placed into different levels of ivs aka Between subjects design or between groups
Attrition
occurs when ppl with extreme scores on prestests drop out of the study, making the overall average lower or higher comparison group similar
measurement error
factors that inflate/deflate person's true score ex person sloutching when measuring height
Random assignment
guards against selection effects all participants have and equal chance of being in each group
Two Independent group designs
post test design only pretest posttest design
Matched-groups design
researchers measure groups on traits that apply to DV, then use those smaller groups to randomly assign participants to levels of IV used when random assignment too random and cant guard against traits that effect dv and evenly distibute participants with this trait amongst groups.
Systematic and unsystematic variability
systematic variability is a confound, its change result in or have an effect on, the outcome unsystematic is something that effects the DV but not consistently can obscure differences in DV but not a confound
