PS291
Terms
undefined, object
copy deck
- equating procedure
- control method whereby participants are assigned to experimental groups by qualities such as gender, race etc.
- matching
- participants are matched in terms of a pretest
- counterbalancing procedure
- participants assigned in a pattern of "ECCEECCE.." Ensures that each condition appears equally often and that each condition follows the other an equal number of times
- randomization
- the compilation of participant assignments solely on the basis of chance
- random sampling
- the selection of participants from a larger population to participate in a piece of research
- random assignment
- the assignment of selected participants to groups
- advantage of randomization over other equating procedures
- randomization controls for both known and unknown potentially confounding variables
- in an outline form, a design tells us..
- what will be done to whom and when
- design must provide a logical structure that enables us to pinpoint the effects of the independent varialble on the dependent variable and thus answer our research questions.
- design must help to rule out confounds as an alternative explanation for our findings
- posttest-only control group design
-
R group.A T M
R group.B T (zero level) M
R=random assignment
T=treatment
M=measurement - pretest
- measurement before the treatment
- posttest
- measurement after the treatment
- pretest-posttest control group design
-
R group.A M T M
R group.B M T (zero level) M - controlled (in reference to some factor in an experiment)
- it is.. both manipulated and constant
- the solomon four-group design
- groups 1 and 2 are a pretest-posttest control group design. groups 3 and 4 are posttest-only control group design. This is to account for pretest influence on results.
- a true experiment includes 3 characteristics and sometimes a fourth. They are..
-
random assignment.
at least 2 levels of independent variable.
controls for major threats to internal validity.
compare two alternative theoretical positions. - between groups variance is influenced by..
- differences between the groups and chance variation
- within-groups variance is influenced mainly by..
- chance variation
- F ratio
-
a statistic for comparing the variance attributed to chance with that attributed to treatment effects, calculated in analysis of variance (ANOVA)
F=t² -if you square the formula for t, you obtain direct measures of variance rather than standard deviations. - analysis of variance (ANOVA)
- an inferential statistical technique used to comparre differences between two or more groups with the purpose of making a decision that the independent variable influenced the dependent variable
- the central limit theorem
- states that if a number of samples are drawn from a population at random, the means of the population tend to be normally distributed
- standard deviation (SD)
- a measure of variability calculated by taking the square root of of the variance
- sum of squares (SS)
- a number used in the determinationof the variance; calculated by summing the squared values for the deviation of each data point from the mean of that data set
- internal validity
- confindence that you can make causal statements about the results of the study - no glowing confounds
- external validity
- confidence that results of study can be generalized to population at large
- construct validity
- confidence that the independent and dependent variables represent the idea behind the study
- variance
-
a description of how much each score varies from the mean.
the average of the squared deviations from the mean.
ss÷Number of scores - between-subjects designs
-
any given participant receives only one level of independent variable.
only one one score for each participant is used in the analysis of the results. - within-subjects designs
-
presents different levels of independent variable to the same group of participants.
every participant serves in every group and receives all levels of the independent variable.
participant's own performance is the basis of comparison - post hoc test
-
similar to performing a t test between the different groups in a multilevel design (1and 2, 2and 3, 1and 3 etc..)
performing a large number of post hoc tests tends to lead to more of them being significant by chance than if you only performed a few tests (familywise error rates) which can lead to type 1 error. - a prioiri tests
- planned comparison tests (to avoid familywise error rates)
- factorial designs
- allow us to examine the effects of more than one independent variable, both individually and collectively, on the dependent variable
- 2X2 factorial design
-
2 levels of factor A and 2 levels of factor B
factor = independent variable
thus 6X2 = 6 levels of factor A and 2 of B - main effects
-
the treatment effect of each independent variable.
the treatment differences (between levels of a given factor) in a factorial design - interaction effect
- the result of 2 independent variables combining to produce a result different from that produced by either independent variable alone.
- cell mean
- influence of the main effect and the interactions
- in factorial design there are many null hypotheses
- one for each factor of the design
- a 2X2 factorial design would have 3 null hypotheses
-
1 to state no differences between the 2 levels of factor A
1 to state no differences between the 2 levels of factor B
1 to state that there are no interaction effects - to assess interaction effects
- determine whether the means of the results pertaining to the two levels of one independent variable changes with the level of the other independent variable.
- subject variable
- a characteristic or condition that a participant is seen to possess in a relatively permanent manner
- within-subjects design equates groups (same participants for each treatment), increses sensitivity by decreasing the chance or error variance because
- it removes the variance that results from individual variability - we expect behaviour of same participant in a series of conditions to be more similar than the behaviour of different participants in a series of conditions
- intrasubject counterbalancing
- ABBA or BAAB order for conditions to help control for effects that are carried over from one trial set to another
- most of the error variance in psychological studies results from differences among participants in the study
- thus in within-subjects designs the F ratio is more sensitive to changes in the treatments
- winthin-subjects design is not appropriate when the treatment has long lasting effect or when the purpose of the study is to test for long lasting effect.
- winthin-subjects is extrememly sensitive to time related effects- fatigue effects or practice effects (called order effects)
- intragroup counterbalancing
- each condition must occur equally often and each condition must precede and follow all other conditions an equal number of times.
- incomplete counterbalancing
- each condition occurs equally often but each condition does not precede and follow all other conditions
- mixed designs
- designs that include both "within" and "between" components. ie having more than one group but using repeated tests of independent variable levels on each group.
- matched-subject design
-
using participants who are very similar in each group.
not as much error variance as between-subjects design or as little as a within-subjects. - for a matching procedure to work there must be a high correlation between the variable used for matching and the independent variable.
- any physical or mental characteristic that can be measured can be used for matching - height weight intelligence anxiety level hair colour
- randomized block design
- the procedure used when you want to analyze the matched factor
- ecology
- the scientific study of the relationship of living organisms with each other and their environment
- ecology of psychological experiment
- examins the relationship among the scientist, the research paticipant, and the experimental situation.
- ecological validity
- seeks to determine whether the impact of the relationships among the scientist, the participant, and the context have been considered fully in evaluating a given piece of research.
- experimenter effects
- portion of the results of an experiment that can be said to be affected by the attitudes or behaviour of the experimenter.
- the personal equation
-
the constant error in the observations of different scientists.
physiological and psychological makeup between scientists causes them to observe the world differently, but consistently. - interrater reliability
- the correlation between two raters.
- subject factors
- present when the research participants are not behaving in the way we expect them to behave.
- Hawthorne effect
- effect of participants reacting to something other than the independent variable
- demand characteristics
- occur when a participants response is influenced more by the research setting than by the independent variable
- closed system
- the important factors that influence the environment are controlled by the experimenter
- archival research
- uses existing records, collected before the time of the study and not for the purpose of the study.
- superordinate goals
- goals strongly desired by members of competing groups, but absolutely not obtainable without the resources and efforts of both groups together.
-
quasi experimental designs
evaluation research designs - less rigorous designs that rely on logically discounting alternative interpretations of data. - lack experimental control and control over potentially confounding variables.
- time series desgn
- within subjects design
- single group pretest posttest design
-
comparing a single pretest measure with a single postest measure with one group
-not knowing the normal amount of fluctuation between any two measures is a weakness of this design (fixed by interrupted times series design) - interrupted time series design
- invovles making several pretest and posttest measurements, leading to a better estimate of the normal fluctuations from test to test. - also can provide some estimate of how long lasting the influence of the phenomenon is.
- multiple time series design
- finding a control group, not directly exposed to phenomenon being studied, to control for other possible interpretations of results. - this makes it a between subjects design
- nonequivalent before-after design
-
used to make comparisons between groups strongly expected to differ in important ways even before the experiment begins.
comparison of pretest and posttest scores of control and experimental groups attempts to control for unequal status. Pretest serves as a "baseline" -
Single subject designs
(small N designs) -
data from a single participant remains separate and not statistically averaged with those of other participants.
assumes that the process under study is found within that single subject and can be controlled appropriately. - point prediction
- in a single-subject design, after repeated procedures, a precise prediction (within limits) about an observed phenomenon.
- Single-case designs offer a powerful method for determining which techniques work best with a given individual.
- oh yeah
- 2 different purposes of single subject designs
-
descriptive - represented by case-study methods. - these are SS naturalistic observation, or the observation of a single variable.
Experimental - focusing on how the introduction of a particular factor influences a particular aspect of an individual's behaviour.
these may be combined
can be used to direct experimental procedures - Naturalistic case studies
-
focus on problematic or exceptional behaviours.
ability to present the clinical implications of a particular disorder.
ability to describe processes not easily reduced to a single variable.
study of individual case in its actual context.
uses patients recollection of events, information from friends, relatives and public records
open to problems of self report - one-shot case studies
-
means of obtaining information about specific interventions in therepy during an ongoing deries of sessions.
can provide valuable initial descriptions of new phenomena, in preparation of more careful rigorous experimental methods.
does not give strong inferences, because of lack of control procedures - Experimental single-subject designs
- intrasubject replication - pattern established with respect to a series of baseline and treatment conditions over time. (pattern repeated with a single participant)
- intersubject replication
- when more than one participant is used in single subject research to see if the data will be replicated between participants.
- reversal design
- shifting the baseline and treatment conditions to observe whether or not the participant is responding to the treatment. ABA type design pattern with as many repetitions as are deemed necessary.
- one limitation of reversal design is..
- it will work only when studying the effect of treatment conditions on behaviours that will return quickly to baseline levels once the treatment is over.
- Multiple-baseline design
-
monitor several behaviours of a single subject simultaneously - establish baseline levels for each behaviour - apply treatment to one behaviour - likelyhood of causal relation is inferred from changes occurring in only the the behaviour exposed to the treatment, non treated behaviours remain the same.
Can be used for behaviours that are permanently changed by the treatment, unlike reversal designs. -
multielement design
alternating treatment design
simultaneous treatment design -
(single subject) comparing different levels of a given variable.
-incorporates many reversals that may be fast paced
-differences in effects of treatments seem to emerge quickly
-best suited to changes that are short lived - open ended question
-
has no fixed answer but allows the respondent to answer in any manner
-does not impose the researchers point of view
-participant may give new info not previously considered
-can pose a problem when trying to analyze data, establishing meaningful patterns -
fixed-alternative question
closed ended question -
limits number of reponses participant can make
-not limited to yes/no - open ended and fixed alternative (closed ended) questions..
-
may be combined within a question ie by adding "other" with a line to fill in option
- may be mixed in same questionaire - funeling
- line of questioning begins as an open ended question and then follows with more specific items (to choose from).
- semantic differential
- 7 point scale, bookended by adjectives such as soft/hard, good/bad etc..
- random response method
-
Method to obtain answers to difficult, potentially embarrasing questions.
using a coin, participant answers yes to any flip of "heads" but truthfully to any flip of "tails". Then use a probability formula to obtain "true" percentages of yes/no to answers. - prbability sampling - predetermined chance of any individual being selected given the particular constraints under study.
- includes: simple random sampling, systematic sampling, stratified random sampling, cluster sampling, multistage sampling.
- simple random sampling
-
(probability sampling)
-every member of the population has a known chance of being selected - systematic sampling
-
(probability sampling)
-list the entire population then choose every Nth person listed
-important that the list contain the entire population in an unbiased order. - stratified random sampling
-
used to ensure the representation of particular groups of people ie race, gender, age.
-divide population according to that category then randomly selct from those divisions. - cluster sampling
- randomly select a certain number of population units (villages or city blocks) then enlist the help of the inhabitants.
- multistage sampling
-
relies on sampling at different stages in the process.
(like narrowing the field as you select ie select schools, then schools from a particular area, etc..) - nonprobability sampling
- includes convenience sampling,quota sampling, snowball sampling
- convenience sampling
-
using people who are available ie. whoever walks by on the street or customers at a snack bar.
-biases the sample toward that situation/location - quota sampling
- sets up a quota, or specific types of people.
- snowball sampling
-
used most often when no list of the population exists.
ie prostitutes..locate a single participant and then ask them to help locate others. -
determining the size of a survey sample - 2 questions
how many people are available and how homogeneous are they for that characteristic - less people needed if they are very homogeneous -
2nd question is: how accurate does the answer need to be
-generally, the larger the population to which you want to generalize, or the greater the variability and the more exact the results need to be - increases the required size of the sample - confidence interval
-
upper and lower bound of the percentage in which the results fall, expressed in points or percentage ie +/-5%
(5% in this case would be called the sampling error)