rsch methods final

two branches of stats

descriptive and inferential

three levels of measurement

numeric, rank order, nominal

types of numeric variables

equal interval, discrete, continuous

equal interval variable

numbers are equal amounts apart from each other

ratio scale

equal interval variable that has an absolute zero

types of frequency distributions

unimodal, bimodal, multimodal, skewed, normal

types of skewed distributions

floor v ceiling (right v left)

floor effect

piles up at bottom (skewed right, positive)

ceiling effect

piles up at top (skewed left, negative)

kurtosis

extent to which frequency distribution deviates from normal curvety

types of central tendency

mean, mode, median

variability statistics

variance and standard deviation

variance formula

sum of squares divided by number of scores

standard deviation formula

square root of variance

sum of squares

sum of ((each score minus the mean) squared)

z score

how many standard deviations a score is from the mean

normal curve standard deviation percentages

34%, 14%, 2%

population parameter symbols

mu, theta

sample statistics

M, SD

probability

expected outcomes divided by total possible outcomes

two interpretations of probability

long run frequency interpretation v subjective

addition rule of probability

when two outcomes are mutually exclusive, the chance of getting either outcome is the sum of each individual probability

multiplication rule of probability

the probability of getting both of two independent outcomes is the product of multiplying both individual probabilities

conditional probability

probability that one event will occur given that the other has already occured

steps of hypothesis testing

research v null hypothesis, comparison distribution, cutoff sample score, sample’s score, reject/ fail to reject null

null hypotheses for one v two tailed

one (mu1 > mu2), two (mu1 = mu2)nu

null hypothesis

if the research hypothesis is false, what would this look like?

comparison distribution

model of normal curve for null hypothesis based on mu,theta

standard cutoff sample scores

1.64/2.33 (one tailed) or 1.96/2.58 (two tailed)

sample score

sample’s raw score converted to z score, compared to the cutoff sample score

implications of failing to reject null

inconclusive, not statistically significant

implications of rejecting null

“supports”, not proves, statistically significant

conventional levels of significance

0.05/ 0.01

distribution of means

normal curve made up of means, more than one individual

comparison distribution features

mean is same as the population mean, standard deviation is standard error (muM, thetaM)

standard error/ SEM

square root of variance of distribution of means, which is the variance of the population divided by the number of samples

shape of distribution of means is normal if…

each sample has 30+ individuals OR the distribution of population of individuals is normal

central limit theorem

the averages of samples means has an approximately normal shape

confidence intervals

interval of what we are confident will include the true population mean

confidence limits

upper and lower values of confidence intervalst

steps for figuring out confidence interval

find standard error, find raw scores for 1.96/2.58 std above/ below the sample mean

analyzing confidence interval

if the null hypothesis mean is not included in the confidence interval, then we are 95/99% we can reject the null/ statistically significant

decision errors

errors made when interpreting the results of a study

type I error

falsely rejecting the null hypothesis: seeing an effect where there isn’t one, p too low, alpha

type II error

falsely failing to reject the null: not seeing a result when there truly is one, p too high, beta

effect size

standardized measure of difference between populations, cohen’s d

effect size formula

mu1 - mu2 over theta (difference between means over standard deviation)

cohen’s d conventions

0.2 small, 0.5 medium, 0.8 large

meta analysis

combining effect sizes from different studies to compare (smaller the range the bigger effect size)

properties of measurement from least to most

identity, magnitude, equal interval, true zero

scales of measurement

nominal, ordinal, interval, ratio

reification of a construct

mistaking a CONSTRUCT for a FACT

construct

idea that is used as assumption as if it is real

inductive reasoning

particular instances to general knowledge

deductive reasoning

general to particular

levels of constraint

naturalistic, case study, correlational, differential, experiment

precision v relevance problem

as we go up in precision/ levels of constraint, the relevance of our studies become less applicable to real life

operational definition

the working definition of how a variable is measured

convergent validity

multiple lines of reasoning draw to the same conclusion

social desirability bias

urge to be “social desirable” biases participants to answer in a certain way, giving inaccurate results

interrater reliability

participants agree with each other

test retest reliability

one individual gets the same results multiple times

internal consistency reliability

several different observations made create the same score

effective range

measurements can be understood well on this scale

scale attenuation effects

restricting range of scale can lead to misinterpreted results, ceiling effect/ floor effect

criterion related validity

validity established by correlating measures with one or more criterion measures

predictive validity

how well something can predict a future event

concurrent validity

how well a instrument/ measure correlates to another instrument/ measure

margin of error (MOE)

maximum difference between sample mean and population mean, 1.96 std

features of a boxplot

75-25th percentile scores, whiskers are max and min, line is median, symbol is mean, circles are outliers

reliability v validity

consistent scores v accurate scores; you can have reliability without validity, but not vice versa

theory

must be testable/ falsifiable

generalizability

how well you can generalize the results of a study to a general population

unobtrusive observer

researcher conducting study without participant knowing by blending into background

participant observer

researcher is a part of the study and influences the environment

measurement reactivity

participants reacting differently to study because they know they are being studied

coding

method of organizing behavior into predetermined codes

experimenter reactivity

experimenter affecting results of the study inadvertently

experimenter bias/ expectancy

experimenter interpreting data/ results of the study differently due to bias

correlational research

studying the relationship between two variables

differential research

comparing two or more groups that differ on preexisting variables

cross sectional studies

studies that take data from groups that are different ages, liable to cohort effect

cohort effect

people from the same cohort/ age group are likely to have similar reactions to certain stimulus due to age, confounding variable that is solved by longitudinal studies

confounding variable

varies similarly to the independent variable and manipulates dependent variable despite experimenter’s direct control

artifact

result of confounding variable

moderator variable

variable that affects relationship between other variables

when is a t test used

when we know the population mean, but not the population variance

degree of freedom

n-1, unbiases sample set

t score formula

T = (M - mu) / Sm

troubling trio

low statistical power, surprising finding, p value under 0.05

repeated measures designs

research method where a person is tested more than one, within subjects design

t test for dependent means

two scores for each person and population variance unknown

difference scores

difference between person’s score on one testing versus another testing

population mean for t test for dependent means

zero— assume that the populations are the same/ no effect

assumptions for t test for single sample and dependent means

normal population, robustness, not skewed

robustness

extent to which hypothesis testing procedure is relatively accurate regardless of assumptions being violated

power

probability that the study will give a statistically significant result IF THE RESEARCH HYPOTHESIS IS TRUE

difference between power and alpha

power- if research hypothesis true, alpha- if null hypothesis true

free random assignment

assigning randomly, assignment of one participant doesn’t affect the assignment of another

randomizing within blocks

use blocks for participants, and fill up each condition before adding more participants