Go on. Explore!

lvl 0 • 0 xp

lvl 1

Home Market Decks Interests

Methods Test 3

memorize.ai (lvl 286)

(0)

48 cards

()

Section 1

Preview this deck

Measures of Central Tendency

Front

1 / 48

GRE

0.0

0 reviews

5			0
4			0
3			0
2			0
1			0

Active users

All-time users

Favorites

Last updated

7 years ago

Date created

Mar 14, 2020

Cards (48)

Section 1

(48 cards)

Measures of Central Tendency

Front

ways to quantify the center of data. Mean, median, mode. You gain information about the population by losing information about individual cases/elements

Back

Inflating Type I error

Front

if alpha=.05 then the probability of Type I error for any test increases with # of NHSTs performed Ptest1 + Ptest2 .05 + .05 = .10 Need to make corrections/adjustments to p-values Also known as issue of multiple tests/multiple comparisons

Back

Between subject design

Front

Pros: no issue of order effects, carry over effects, demand characteristics Cons: Need a larger sample size for each group Less sensitive in detecting potential differences More error/unaccounted for variance - subjects don't act as their own control

Back

Interval

Front

Can be measured along a numeric continuum, with consistent meaning of intervals between values. NO meaningful zero point. EX: Fahrenheit or Celsius

Back

Dr. Stephen Heard

Front

"Every statistical test works the same way"

Back

Type I error

Front

rejecting the null hypothesis when the null hypothesis is true.

Back

Test Statistic

Front

measures strength or pattern in our data. Always computed from sample data. Represent the signal-to-noise ratio

Back

Nominal

Front

two or more categories into which people/elements/data points can be placed. CANNOT be ordered or ranked. EX: Political ideology

Back

Ratio

Front

measured along a numeric continuum, with consistent meaning of intervals between values. DOES have a zero point indicating complete absence of a quantity. EX: height and weight, Kelvin

Back

Skew

Front

mean is more affected by extreme values than median, but this is mitigated by sample size

Back

T-Tests and ANOVA

Front

determine whether there are mean differences

Back

Dichotomous AKA Discrete, Qualitative

Front

Only 2 response options are possible. EX: Yes/No questions indicating presence vs. absence

Back

Categorical Family

Front

to measure associations among categorical variables. EX: odds ratio

Back

Inferential Statistics

Front

subjecting data to hypothesis testing

Back

Population

Front

the total group or set about which we want to make inferences and conclusions

Back

Quasi Experiment

Front

no random assignment, but workarounds (pre/post test measures to establish baseline)

Back

Ordinal

Front

two or more categories into which people/elements/data points can be placed. These can be ordered or ranked. EX: Strongly approve, somewhat approve, somewhat disapprove...

Back

The two steps of inferential statistics

Front

1. measure the strength of pattern in our data 2. ask ourselves, is this pattern strong enough to be believed?

Back

Correlation/Regression

Front

Measuring strength of association/slope of a line

Back

Measures of Spread

Front

How variable are the data around the center. How spread out/apart vs close together the data are

Back

Type II error

Front

not rejecting the null hypothesis when the null hypothesis is false

Back

counter balancing

Front

all possible orders of IV are presented across participants

Back

What is the spread?

Front

noise

Back

Within subjects design AKA repeated measures design

Front

everyone receives multiple experimental conditions. Pros: fewer subjects needed more sensitive in detecting potential differences - subjects = groups so less error More ecologically valid Cons: Carry over effect - effects of one IV level can bleed into another Demand characteristics - subjects exposed to all conditions may figure out what hypotheses or questions are being tested

Back

What to do with test statistics?

Front

compare the test statistic against a hypothetical distribution assuming no effect

Back

Strength of an effect?

Front

Effect size

Back

Order Effect

Front

the order of presenting levels/conditions of the IV effects the DV

Back

Whether an effect exists? Is it strong enough to be believed?

Front

NHST

Back

Uncertainty about effect?

Front

Confidence Interval

Back

Correlation Family

Front

to measure associations among quantitative variables; effect size based on variance explained. EX: Pearson's R

Back

Difference Family

Front

to measure mean differences between groups; effect size based on ratio of mean difference to spread. EX: Cohen's d Continuous measure so it can take any value Small = .2 Medium = .5 Large = .8

Back

Less Convincing Patterns

Front

Smaller mean differences, larger variability/spread

Back

T value

Front

Mean difference (signal) divided by variability/standard error (noise)

Back

Effect Size

Front

helps quantify the strength of an effect of association. Another way to capture the signal-to-noise ration in the observed data. The larger the effect size, the better. Different types depending on variables/levels of measurement

Back

Sample

Front

a subset of the total group or set, about which we want to make inferences and conclusions. We recruit and enroll the sample in a study and compute sample statistics

Back

The problem with data

Front

we see apparent pattern, but we aren't sure if we should believe it's real, because our data are noisy

Back

Descriptive Statistics

Front

merely describing data

Back

Confidence Intervals

Front

help quantify uncertainty about true/underlying population parameters we hope we can estimate from the sample. If repeatedly took additional samples from the population, and calculated the confidence interval for each sample, then we would expect the true population parameter to appear in those intervals 95% of the time. NOT the 95% chance the population parameter falls within the one confidence interval calculated from the current sample

Back

P Value

Front

the probability that we would see this data pattern, given the null hypothesis. Retain if the p-value is high or reject if the p-value is low

Back

Variance

Front

The average squared deviation of values from their mean

Back

What is the center of the distribution?

Front

Signal

Back

95% CI = sample mean plus or minus margin of error (critical value times (sample standard deviation divided by square root of sample size))

Front

Want margin of error as small as possible sample standard deviation = as small as possible Sample size = as large as possible critical value = trade off: as confidence increase, interval will increase

Back

Ways to increase the T-value

Front

Numerator (signal)- Increase mean difference Denominator (noise) - decrease variability, remove confounds increase sample size to reduce influence of outliers

Back

Continuous Variable

Front

AKA Quantitative variables

Back

Null Hypothesis Significance Testing (NHST)

Front

Workhorse of inferential statistics. Quantifying what the "surprise factor" of a pattern in the data is, under a certain assumption/condition. The probability that we would see this data pattern, given the null hypothesis.

Back

More Convincing patterns

Front

Larger mean differences, smaller variability/spread

Back

Field experiment

Front

random assignment, but DVs and other measures are assessed in the wild. Disadvantage = less experimental control Advantage = greater ecological validity

Back

Standard Deviation

Front

the square root of the variance. The larger the SD the more spread out the data

Back