Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
51 Cards in this Set
- Front
- Back
ONE SAMPLE T-TEST |
This is a method use to compare the SAMPLE MEAN to the POPULATION MEAN Sample mean will be compared to ZERO. Used to ensure that any difference is due to the SAMPLING ERROR rather than being non-representative of the population An example where a ONE SAMPLE T-TEST would be used is when testing whether the population is OVER OPTIMISTIC |
|
PAIRED/REPEATED T TEST |
LIKELY TO HAVE LOWEST SAMPLING ERROR Comparing Means of TWO SAMPLES - we want the difference to be MORE than zero PAIRED: when the two samples are matched based on characteristics e.g. age, gender, ethnicity - TWINS would be the BEST REPEATED - same participants in two different conditions |
|
INDEPENDENT T-TEST |
Comparing means between TWO INDEPENDENT SAMPLES e.g. one sample in one condition and another in another condition LIKELY TO HAVE THE MOST SAMPLING ERROR due to INDIVIDUAL DIFFERENCES. USE LEVINE'S TEST FOR INDEPENDENT T-TEST |
|
LEVINE'S TEST OF SIGNIFICANCE |
DIAGNOSTIC TEST used When doing an Independent T-test, it will be valid if the VARIANCES ARE ASSUMED TO BE THE SAME. If Levine's test is SIGNIFICANT then we must assume that the VARIANCES are NOT EQUAL - use another D.F |
|
T-TEST |
This is a method to test the difference between means of 2 variables |
|
SAMPLING ERROR |
This is the error of variance which is caused by characteristics of the sample. More likely for SMALL samples |
|
STANDARD ERROR |
Standard deviation of the means of different samples in comparison to the mean of ALL samples. This is because every time a test is conducted on a group, this will lead to slightly different results. |
|
VARIANCE |
A measure of dispersion. How much a result deviates from the mean Squared distance of the mean of each point LARGE VARIANCE can be OFFSET BY LARGE SAMPLES
|
|
ANALYSIS OF VARIANCE (ANOVA) |
Similar to T-test, compares the difference between means MORE THAN 2 variables can be analysed unlike a T-test. This is more likely to be used in complex experiments wanting to find interactions between variables Compares VARIANCE CAUSED BY INDEPENDENT GROUPS to VARIANCE CAUSED BY SAMPLING ERROR |
|
ONE WAY ANOVA |
This is THE SAME AS A T-TEST as there is ONE IV IV can have DIFFERENT NUMBER OF LEVELS and the dependent variable will be measured This is MORE ACCURATE AND PRECISE than T-test Example Testing effectiveness of a drug on a psychological disorder IV : DRUG 2 levels will be the different types of drug e.g. PROZAC and BENZODIAZAPINES DV: symptoms |
|
TWO WAY ANOVA |
Looking at the effects of TWO INDEPENDENT VARIABLES at different levels For example: IV 1 : DRUG : prozac v. benzo. IV 2 : Activity: Social v. non-social DV : symptoms CAN HAVE AN INTERACTION WITHOUT MAIN EFFECT RESULTS: MAIN EFFECTS of EACH IV INTERACTION BETWEEN LEVELS OF IV e.g. less symptoms when SOCIAL and on PROZAC |
|
THREE WAY ANOVA |
3 IV'S and with different levels Can have 2 WAY INTERACTION or 3 WAY INTERACTION MAIN EFFECTS and INTERACTION EFFECTS need to use MAUCHLY test of SPHRECITY as it is testing the DIFFERENCES and there is MORE THAN ONE DIFFERENCE whereas in TWO WAY ANOVA - only one difference |
|
MAUCHLY TEST of SPHRECITY |
Comparing differences ASSUME THAT THE DIFFERENCES ARE EQUAL If SIGNIFICANT, this means that the differences ARE NOT EQUAL --> we must use GREENHOUSE GEISSER measure |
|
BETWEEN SUBJECTS |
When different conditions are assigned to DIFFERENT groups rather than the same person Used when it is not possible to use one person For example, PERSONALITY TRAIT - NEUROTICISM |
|
WITHIN SUBJECTS |
When the same participant/group is allocated to all conditions E.g. same patients given a different drug Can be used when MOOD has to be elicited e.g. someone in a HAPPY MOOD and a SAD MOOD (better to test in the same person) |
|
MIXED SUBJECTS DESIGN |
ONE VARIABLE = BETWEEN e.g. the drugs administered ONE VARIABLE = WITHIN e.g. social v |
|
POST HOC/PLANNED COMPARISONS |
ANOVA can tell us that there is a DIFFERENCE within conditions at different levels but doesn't tell us WHERE the difference is exactly PLANNED COMPARISON - when you have an idea what the effect of the differences may be POST HOC - exploratory |
|
EFFECT SIZE |
The power of Significance Larger effect size = BETTER Effect size = d in T-test Calculated as DIFFERENCE or MEAN divided by the STANDARD DEVIATION In ANOVA = partial eta squared CANNOT BE NEGATIVE |
|
QUALITATIVE DATA |
This aims to be DESCRIPTIVE and provide WEALTH of information that is INDEPTH aims to DESCRIBE, DECODE, TRANSLATE and come up with MEANING for phenomena 1) doesn't treat things equally if they are similar as with QUANTITATIVE DATA 2) doesn't RANK |
|
CONVERSATIONAL ANALYSIS |
Use of TRANSCRIPTS - can identify things embedded in speech and interpret them to indicate implicit attitudes or feelings etc. |
|
DISCOURSE ANALYSIS |
SOCIAL CONSTRUCTIONISM TWO STRANDS IDEOLOGICAL (historical functions and cultural resources) RHETORICAL (immediate functions of talk) |
|
GROUNDED THEORY |
GLASER & STRAUSS (1967) Aims to generate NEW IDEAS - DISCOVERY STRAND of quantitative data EXAMINATION and RE-EXAMINATION of cases and EXTRACT STRUCTURE Process DATA COLLECTION CODING CORE ANALYSIS: Refine indexing system, Memo writing, Category linking OUTCOMES : Key concepys, Definitions, Memos and Relationships/models |
|
CORRELATION |
Finding the RELATIONSHIP between two CONTINUOUS variables we can see how one variable may influence another Correlation variable R is between -1 and +1 Number indicates the STRENGTH of the relationship POSITIVE CORRELATION - when variable 1 goes up, variable 2 goes up NEGATIVE CORRELATION - when variable 1 goes up, variable 2 goes down NO CORRELATION - no relationship between the variables |
|
SCATTER DIAGRAM |
Illustrates the relationship between the two variables each variable plotted on EACH AXIS -useful for identifying OUTLIERS LINE OF BEST FIT can be produced |
|
PEARSONS' PRODUCT MOMENT CORRELATION |
Measure of correlation - LINEAR relationship between X and Y |
|
PARTIAL CORRELATION |
Correlation between TWO VARIABLES can be influenced by a THIRD VARIABLE acting on BOTH IN PARALLEL for example, VIOLENT TV correlates with AGGRESSION as they are both influenced by PARENTAL DISAPPROVAL can CONTROL for the THIRD VARIABLE and make CONSTANT mathematically or experimentally Can see whether or not the two variables STILL correlate when this variable is constant |
|
PARTIAL CORRELATION 2 |
LOWER PARTIAL CORRELATION than ZERO ORDER correlation - the third variable DOES have an effect e.g. could be the reason WHY two variables correlate NO PARTIAL CORRELATION/THE SAME CORRELATION as before - the third variable has NO EFFECT - the TWO VARIABLES CORRELATE HIGHER PARTIAL CORRELATION - this suggests that the influence of the THIRD VARIABLE DIFFERS to each variable e.g. may correlate NEGATIVELY on one variable and POSITIVELY on another |
|
SUPPRESSION EFFECT |
This is when a THIRD VARIABLE has a negative effect on ONE VARIABLE and a positive effect on ANOTHER VARIABLE - leads to SUPPRESSION of one variable EXAMPLE Being with friends as a THIRD VARIABLE ENHANCES AGGRESSION DECREASES THE AMOUNT OF TV WATCHED |
|
REGRESSION |
UNDERLYING ASSUMPTION that there is a LINEAR RELATIONSHIP between variables A formula whereby a dependent variable (outcome) can be predicted by predictors Uses the relationship between X AND Y In the format: y = bX + c b - c - the intercept |
|
REPEATED MEASURES TWO WAY ANOVA |
When BOTH IVs are manipulated in the SAME INDIVIDUAL e.g. DRUG TAKEN and STRESS LEVELS |
|
ETHICS |
MORAL PRINCIPLES These are the guidelines that any researcher must follow for their experiments; taking morality into account for their research - BPS GUIDELINES MINIMISING HARM TO PARTICIPANTS - e.g. preventing physical and psychological stress INFORMED CONSENT - includes right to withdraw DECEPTION DEBRIEFING Studies such as MILGRAM led to a lot of criticism being accused of not being humane |
|
MINIMISING HARM |
Researchers must not INTENTIONALLY cause any psychological harm/distress or physical harm to participants, No INVASION OF PRIVACY or DIGNITY If participants appear to be suffering, they should be able to withdraw If there harm DOES occur, researcher should be prepared to compensate e.g. providing counselling (part of debriefing) |
|
INFORMED CONSENT |
Participants must be given a document which states WHAT they will be doing in the experiment. Must detail to appropriate level anything If there may be anything that causes HARM or DISCOMFORT MUST be detailed Must include the RIGHT TO WITHDRAW at ANY POINT and to DISCARD THEIR RESULTS Must be SIGNED by participant Must be given a copy of their own Understanding that there may need to be a bit of decept in order for correct results to be collected - if this is the case, they MUST debrief after the experiment |
|
DECEPTION |
Researcher should not go out to initially or maliciously LIE to the participant Understanding that dropping a FEW details that will keep participants blind may be needed a FULL DEBRIEF should follow this |
|
DEBRIEF |
After the experiment, researchers should EXPLAIN THE REAL purpose/aim of the experiment to the participants May be required face to face AND in writing Offered counselling if necessary |
|
MULTIPLE REGRESSION |
This is when you find an outcome based on MULTIPLE PREDICTORS ASSUMES LINEAR RELATIONSHIP BETWEEN PREDICTORS AND OUTCOME For example - negative affectivity, negative life events will predict well being Types: SIMULTANEOUS STEPWISE HIERARCHICAL |
|
SIMULTANEOUS MULTIPLE REGRESSION |
ENTERING ALL PREDICTORS ASSESSING THEIR EFFECTS ON EXPLAINING VARIANCE OF THE OUTCOME INDEPENDENTLY - partials out the other predictors For example, to test outcome of well being - enter LIFE EVENTS and then SLEEP QUALITY Beta weights - tells us PREDICTIVE POWER of each variable COMPARING PREDICTORS |
|
STEPWISE MULTIPLE REGRESSION |
SPSS decides what the order of predictors will be. EXPLORATORY METHOD Tries to find the BEST combination of IV that predicts the outcome Begins with ENTERING THE MOST INFLUENTIAL IV which predicts the MOST variance NEXT which predicts the NEXT MOST VARIANCE and so on until adding an another IV will not have any effect on explaining variance PREVIOUS IVs can be taken out if they no longer predict outcome once other IVs have been added MAXIMISING PREDICTIVE POWER |
|
HIERARCHICAL MULTIPLE REGRESSION |
When you decide which IVs are entered FIRST based on THEORETICAL RELEVANCE. Usually enter GENDER/AGE first followed by PERSONALITY followed by STATE TEST FOR SPURIOUS EFFECT CONTROL FOR effects of PREVIOUS PREDICTORS |
|
LIMITATIONS OF STEPWISE REGRESSION |
May not have THEORETICAL importance |
|
SPURIOUS EFFECT |
This is when TWO VARIABLES appear to correlate but only because a THIRD VARIABLE is influencing them in PARALLEL Therefore, in Multiple regression, when adding this variable, it is likely that the variable that PREVIOUSLY appeared to predict the outcome will NO LONGER E.g. NEGATIVE AFFECTIVITY --> influence WELL BEING AND HASSLES Hassles was at FIRST a predictor of well being NA becomes significant predictor when controlling for Hassles - which no longer is a significant predictor |
|
REQUIREMENTS FOR MULTIPLE REGRESSION |
IV and DV must have LINEAR RELATIONSHIPS DV must be NORMALLY DISTRIBUTED IV must NOT be MULTICOLLINEAR (must not correlate with one another) OUTLIERS MUST BE REMOVED |
|
MEDIATOR |
This is when there is a factor that intervenes between a PREDICTOR and an OUTCOME Explains HOW a predictor leads to an outcome 3 ways to check whether a variable is a mediator: The predictor must predict the outcome The Predictor must predict the mediator The mediator must predict the outcome The relationship between the predictor and the outcome must be reduced when the mediator is controlled for CAN BE CONFUSED WITH SPURIOUS EFFECT To disambiguate, MEDIATOR = TEMPORAL SEQUENCE The mediator occurs PRIOR to the OUTCOME Example: ATTITUDE --> INTENTION --> BEHAVIOUR (The theory of planned action, Ajzen and Fishbein, 1975) |
|
MODERATOR |
This is a variable which CHANGES the relationship between the predictor and the outcome e.g. the strength of the relationship This relates to WHEN the predictor predicts and outcome Can be worked out by: Multiplying the predictor and the suspected moderating variable = CROSS PRODUCT Standardise Test whether it accounts for any results after the main effects of the variables has been tested Example ATTITUDE --> ATTITUDE STRENGTH (Fazio 1984) --> BEHAVIOUR |
|
FACTOR ANALYSIS |
This is a DATA REDUCTION method which is used to find the underlying dimensions in a correlation matrix with a large number of correlations. Used to find dimensions of PSYCHOLOGICAL DISORDERS e.g. OCD and INTELLIGENCE Different types include: Exploratory Confirmatory |
|
EXPLORATORY FACTOR ANALYSIS |
When you are assessing the factorial structure of data with NO PRIOR ASSUMPTION Most likely to be used |
|
CONFIRMATORY ANALYSIS |
When you are confirming the factorial structure of data that is pre-defined |
|
EXTRACTION OF FACTORS |
Use of PRINCIPAL COMPONENTS ANALYSIS Method used to find the factors which account for the variance in the data First component = the MOST variance Second = the most of the remaining and so on until the variance is fully accounted for This can lead to MANY components |
|
FACTOR RETENTION |
Methods to decide which factors to retain: 1) KAISER CRITERION use of eigenvalues - value which accounts for the amount of variance that each compenents account for Eigenvalues > 1 must be retained 2) SCREE PLOT retain until the elbow in the graph 3) INTERPRETATION retain factors which will be theoretically relevant to interpret |
|
ROTATION |
Method of rotating factors to make sure that variables are more correlated/not correlated with the factor --> leads to EASIER INTERPRETATION ORTHOGONAL - factors CANNOT intercorrelate e.g. Costa and McCrae big 5; Eysenck 3 personalities Use of VARIMAX --> easier to interpret OBLIQUE - factors can CORRELATE Use of OBLIMIN CATTELLs 16 personality types |
|
REQUIREMENTS FOR FACTOR ANALYSIS |
Relationship between predictors and outcomes must be linear Outliers must be removed Sample size must be GREATER than 100 Normality |