Help
Options
Didn't know it?
click below
click below
Knew it?
click below
click below
Don't Know
Remaining cards (0)
Know
retry
shuffle
restart
0:00
Embed Code - If you would like this activity on your web page, copy the script below and paste it into your web page.
Normal Size Small Size show me how
Normal Size Small Size show me how
SARD psych 2a
| Question | Answer |
|---|---|
| psychological construct | theoretical idea in psychology, not directly observable |
| operationalisation | process of turning construct into something objectively measurable |
| experimental research | establish cause + effect relationship, controlled environment, control group, random assignment to conditions, IV manipulated to measure effect on DV |
| within-subject | study's participants put through all levels of experiment |
| btwn-subject | randomly assigned groups only go through one level of experiment |
| non-experimental design | association established btwn two variables, observation, limited control of external variables |
| correlational design | associations/relationships established btwn two variables, can do statistical correlation test, can be continuous/categorical variables |
| cross-sectional design | different cohorts (usually age groups) measured at one single time |
| cohort effects | variables affecting diff cohorts differently due to their subjective temporal experience (eg generational differences), making cohorts less accurately comparable |
| longitudinal design | one cohort/sample studied over extended period |
| cross sequential | following different cohorts of age in closer succession, combo of cross-sectional and longitudinal elements |
| quasi experimental design | experimental design without the ability to randomly allocate participant groups, no control group, little control over extraneous variables |
| post-test-only design | intervention introduced and results measured after |
| pre-test, post-test design | baseline measured, intervention introduced, results measured after |
| non-equivalent pre-test, post-test design | baseline, intervention, remeasure for multiple groups |
| post-test only eval | + = quick to implement - = don't know participants baseline, cant compare disposition before and after intervention |
| pre-test-post-test eval | + = insight into baseline, quick - = repetition of test may lead into practice effects, history effects, regression to mean |
| regression to mean | extreme values = closer to mean upon second test |
| history effect | events happening outside experiment = affect how partips respond in study |
| practice effect | participants scores improve/change due to prev exposure to test, not own development |
| non-equivalent eval | + = baseline comparison, quick, most informative - = no random assignment, regression to mean |
| sample (n) | smaller portion of population |
| population (N) | all possible membs of category from which sample is drawn |
| representative sample | sample which accurately represents wider pop |
| generalisability | extent to which results of study can be applied across populations |
| probability sample | sampling techniques where each member of pop has equal probability of being selected, need access to all membs of pop |
| simple random sampling | pulling names from hats, random number generator |
| stratified sampling | group pop into stratas/groups based on characteristics, random sampling from stratas (proportional vs disproportional) |
| stratified disproportional | each strata represented equally, might not reflect proportions in larger population |
| stratified proportional | each strata represented to reflect proportions of original pop |
| cluster sampling | putting pop into clusters, use cluster as sample (eg one school to represent all schools of area) |
| systematic sampling | use of sampling interval (k), choose partips based on randominterval repetitively, must choose partip in interval otherwise not probability sampling |
| sampling interval | k = N/n |
| non probability sampling | not all in pop have same prob of being selected, sampling from researchers' judgement, biased sample |
| convenience sampling | choosing partips which are convenient based on geographical location, availability, volunteer sample |
| purposive sampling | sampling based on certain characteristic or experience of desired sample |
| quota sampling | same as stratified without access to whole pop, proportional and nonproportional |
| snowball sampling | network sampling request those successfully sampled to recruit from own social networks |
| self selection bias | unable to reach those who do not volunteer themselves |
| non-coverage bias | not covering all bases of pop,,eg only reach ppl with laptop access in an online study |
| null hypothesis significance testing (NHST) | testing whether signif results happened by chance or if results reflet actual phenomenon in society, (t test, chi sqaure, ANOVA, correlational) |
| descriptive stats | summarise + organise data (mean, median, mode etc) |
| inferential stats | allow to make inferences abt wider pop/phenomenon |
| p value | critical value, cut off point deciding if results significant or not, usually 5%/0.05 (arbitrary) |
| alpha | a, 5%, p value |
| beta | B, 20% |
| power | 1 - Beta |
| effect size | quantifies size of effect of IV on DV |
| sample size | number of ppl participating in study/supplying data |
| statistically significant | can assume results did not happen by chance + reflect phenomenon of wider pop |
| type 1 error/false positive | falsely rejecting null hypothesis |
| type 2 error/false negative | falsely failing to reject null hyp |
| point estimate | best guess of population parameter |
| confidence interval | plausible range of values of population parameter |
| variable types | nominal, ordinal, interval, ratio |
| nominal | binary responses |
| ordinal | ordered scale, undefined interval btwn values(likert scale) |
| interval | no true zero, equal intervals btwn values |
| ratio | interval with true zero, no values below zero |
| Cross-Tabulation Chi-Square Test | describe relashe btwn 2 categorical variables, know how one variable associated w distribution of outcomes in another variable (diff btwn proportions of 2 variables), determine expected frequencies, each partip only come up once |
| One-sample t-test | test if mean of random sample differs from known mean of larger pop (H0=means are equal btwn sample and pop) |
| One-sample Wilcoxon signed-rank test | non-parametric equivalent of one-sample t-test, compares median of sample against a single value |
| Residuals | difference btwn actual dat point and sample mean |
| one-sample t test assumptions | 1. continuous DV (interval or ratio data) 2. dat is independent (no relashe btwn observations) 3. no extreme outliers 4. normality (residuals normally distributed |
| Effect size | strength of relationship shown, Cramer's V |
| chi-square test assumptions | 1. dat is categorical 2. observations are independent (value of one obs does not affect value of another) 3. cells in table are mutually exclusive (each dat point belongs to only one cell in table) 4. expected frequencies are sufficiently large |
| paired sample t test | - testing if significant diff btwn 2 groups of related data - within participant design - data taken before and after intervention |
| paired sample t test other names | - repeated measures t test - matched sample t test - within subject t test |
| mu | population mean |
| paired t test comp to one sample t test | paired t test = replace sample mean w average of diffs of paired scores + set pop mean to 0 - a paired t test is like a onesample t test on the diffs btwn pairs of scores, assuming pop mean is 0 |
| paired sample degrees of freedom | df = n -1 |
| assumptions of paired sample t test | - dv = interval or ratio (continuous) - iv = nominal or ordinal w exactly 2 levels (categorical) - dat is independent, no missing values (2 from each partip, each pair from diff partips) - normality (residuals are normally dist) |
| residuals calculation | residuals = difference score - mean difference |
| paired sample t test procedure | 1. sum up all results for both time points 2. calc mean + sd for each time point 3. calc difference scores for each partip, cal mean + sd of diff scores 4. plot figure (scores at diff time points or difference scores) 5. check assumptions |
| paired test step 6. compute t test w formula | if t > critical value = significant if t < critical value = not signif |
| paired t test step 7. calc standardised effect size | - cohen's d = mean of differences/sd of differences - interp -> 0.20 = small, 0.60 = med, 0.80 = large effect |
| non parametric equivalent of paired t test | - applies if any assumptions violated - Wilcoxon signed-rank test (dat needs to be ordinal, interval or ratio, needs ind. scores) |
| wilcoxon signed-rank test proc | - data ranked first (absolute value) - nhst applied by comp observed ranks associated w pos + neg differences - H0=assume ranks evenly dist around 0 - signif result = ranks not evently dist around 0 |
| two sample t-test | - tests if signif difference btwn 2 groups of unrelated data - compare two groups/conditions where partips are diff in each group (between groups design) |
| other names for two sample t-test | - independent sample t-test - btwn subject t-test - unpaired/unrelated t-test |
| 2 versions of two sample t test | - student t test - welch t test - difference = assumption of homogeneity |
| non parametric two sample t test | Mann-Whitney U-test |
| two sample t test assumptions | - continuous dv and categorical iv (with 2 lvls) - independent dat (partips only belong to 1 grp) - homogeneity of variance (student t-test only) = aka homoscedasticity, variance btwn groups are equal - normality (residuals = normally dist) |
| variance | sd^2 |
| welch t test | assumes by default that variances are unequal - unequal variance = heteroscedasticity - recommended, as gives same results when variances are equal as student and more accurate results when variances are unequal |
| welch t test df | - tend to have decimals (how to distinguish btwn welch + student t test) |
| correlational research | - non experimental research, find if 2 variables realted to each other - little/no control over extraneous cariables - if two variables associated = correlated = covary - correlation IS NOT causation |
| 3rd variable problem | unmeasured/unintended variable (may have effect, unable to control) |
| characteristics of correlational relationship | - strength - direction |
| direction of correlation | - pos or neg |
| positive correlation | - variables move in same direction - one goes up/down, so does other |
| negative correlation | - variables move in opposite direction - one goes up other goes down, vice versa |
| strength of correlation | - variance/variability = quality of idff in scores of variables |
| covariance | - measure of how 2 variables vary/change together - if covariance is positive = positive relashe |
| calculating covariance (description of formula) | 1. for each partip, subtract their value from mean for both variables, mult tgthr to get one value for each partip 2. sum all of values from step 1 3. divide by numb of obs minus 1 |
| problem w covariance as measure of relashe | - sensitive to units of measurement of variables (covariance may be large becuase units are large, not because relashe is strong) - cov in hrs - 87.78, converted to mins = 316,008 - solush = standardised covariance |
| standardised covariance | - divide covariance by product of stand dev of both variables - get correlation coefficient = r(x,y) |
| correlation coefficient | - numerical value ranging from -1.00 to 1.00 - closer to -1 or 1 = stronger correlation - 1/-1 = perfect neg/pos relashe - strong = 0.7-9 - moderate = 0.4-0.6 - weak = -0.1-0.3 - 0 = no relashe |
| parametric correlational analysis | Pearson's product moment correlation |
| assumptions of pearsons product moment | 1. both variables = continuous dat (interval/ratio) 2. related pairs = 2 dat points from each partip 3. ind of obs 4. lintearity = relashe btwn variables is linear 5. normality of residuaks 6. homoscedasticity 7. no outliers |
| linearity assumption for pearsons | residuals vs fitted plot - residuals = y axis - fitted (predicted) valuyes = x axis - lookinfg for roughly flat horizontal line |
| normality of residuals assumption pearsons | Q-Q plot - majority of values should fall on/close to diagnoal red line |
| homoscedasticity for pearsons | - variability/spread of one variable remains constant across range of another variable scale-location plot - roughly random spread of points as move from one end to other - red line roughly flat + horizontal |
| visualise relashe btwn 2 continuous variables | scatterplot |
| non parametric correlation test | spearman's correlation coefficient - for when one of pearsons assumptions = violated (eg one variable = ordinal dat) |
| spearman's correlatin coefficient | - calc relashe based on rank order of dat, rather than actual values |
| spearmans proc | 1. rank scores for each var separately 2. calc diffs btwn ranked pairs 3. square diffs 4. sum squared diffs 5. complete spearman's formula |
| tied observations | - when 2+ obs/partips have same value - Spearman's NOT sued w tied ranks - when obs appears more than 1ce = sum of natural ranks/numb of times value apprs in data set |
| tied observations pearsons | 1. subtract observations from mean, multiply to get one value for each obs 2. sum all values from step 1 3. diide by numb of obs |
| comparing two correlations | eg: comp whether correlation self esteem + depression is significantly diff from correlation btwn self esteem + anxiety - dependent or indep correlations |
| dependent correlations | - share common variable, come from same sample |
| independent correlations | - same variables, diff groups - comparing corr btwn study hours + statistical anxiety in students who like maths and DO NOT like maths |
| correlation vs regression | corr - how strongly two vars relate to each oth (strength + direc) regresh - determines if 1 var predicts/expl another var |
| predictor variable aka | - independent var - explanatory var - x variable (in regression model) |
| outcome variable aka | - dependent var - criterion var - y variable |
| simple linear regression | one predictor var predicting one outcome variable |
| multiple linear regression | two or more predictor variables predicting one outcome var |
| H0 phrasing linear regression | *predictor var* does not significantly predict *outcome var* |
| mean model | - uses average of all outcome scores - regardless of predic var, predict outcome var as same/mean - not good model. does not capture dat well |
| line of best fit model | - ajudst predicted value based on x - line of best fit = regresh line - predicted y score for each x |
| residual | - diff btnw regresh line/predicted score vs actual score formula = actual value-pedicted value aka error (smaller dist btwn actual val + regresh line = better predic, lower error) |
| positive residual | - actual value ABOVE regresh line - model underestimates outcome (y) var based on predictor (x) var - actual y is higher than predicted - y is more likely than predicted |
| negative residual | - actual val BELOW regresh line - model overestimates outcome (y) var based on predictor (x) var - actual y is lower than expected - y is less likely than predicted |
| simple linear regression equation | Yhat = b0 + b1X Yhat = predicted outcome based on regresh mod X = predictor var b0 (betea-zero) = intercept, predicted Y when x=0 b1 (beta-one) = slope, change in predicted Y for each 1 unit increase in X |
| assumptions of simple linear regresh | 1. outcome var = continuous 2. predictor var = continuous or 2 lvl categorical 3. ind obs (each score from diff partip) 4. non zero variance 5. linearity (linear relash btwn vars) 6. normality of residuals (residuals = norm dist) 7. homoscedas |
| non zero variance | - predic var must var - not all values same across scores (need spread) - lack of variability (everyone w same score) = imposs to examine var relashe |
| homoscedasticity of simple linear regresh | - variance of residual is constant along values of predictor variable - check w scale-location plot, testing hyp of (non-) constant error, Breusch-pagan test |
| analysing simplie linear regresh results | R^2 F-test Regression coefficient (b) |
| r- squared | - coefficient of determination - indicator of goodness of fit - proporsh of variability in outcome variable that is explained by the predictor variable - varies btwn 0 and 1 |
| R^2 = 0.00 | - no variance of outcome var explained by predic variable |
| R^2 = 1.00 | - perfect prediction (pred var perf predicts outcome var - all variance in outcome var explained by predictor var |
| R^2 = 0.8 | 80% of variance in outcome var explained by predictor var - remaining 20% of variance may be due to extraneous vars not incl in model |
| F-test | tells if R^2 + regresh model as whole is statis signif - explains stat signif protion of variation in outcome var |
| relashe btwn R^2 and F-test | R^2 = how much of variance is explained F - test = whether what is explained is stat signif (or occurring by chance) |
| adjusted R^2 | - adjusts R^2 for numb of predictors in model - provides more acc measure of goodness of fit - tells us more abt model acc when more than one predic var is used |
| unstandardised regression coefficient | b - slope if b = 0.30, w every 1 unit incr of predictor var expect outcome var to increase by 0.30 |
| degrees of freedom simple linear regression | df1(regression df) = numb of predictors - 1 df2(residual df) sample size - numb of predictors - 1 |
| effect size for simple linear regression | cohen's f^2 |
| pos regresh coefficient | predict + outcome var move in same direction - b = 2,5 |
| neg regresh coeff | predict + outcome vars move in opp directions - b = -0.80 |
Created by:
melissa.sjolin