LING 333 Quiz #3

Cohort Effect

each generation acquired a different variant and keeps it as they age

Experimental Design Factors (3)

• researcher manipulates the independent variable

• participants are randomly assigned to conditions

• can establish a causal relationship

What is the purpose of experimental design

desire to establish a relationship between both variable

Manipulation

must manipulate and decide what goes in the levels of the independent variable

Random Assignment

the condition the participants receive must be random

Non experimental

• researcher measures or compares what already exists

  • everything is pre existing nothing is randomly assigned

• no random assignment to conditions

• can show association but not causation

When should you choose an experimental design?

when the intention is to establish a causal relationship

you have a hypothesis about a causal relationship between two variables

you are able to manipulate the independent variable

you can randomly assign participants to levels of the independent variable

all three must hold missing one → non-experimental

Quasi-experimental designs definition

the researcher manipulates the IV (delivers a treatment) but participants are NOT randomly assigned to conditions (if its impractical or unethical)

why do we have quasi-experimental designs

when random assignment is impractical or unethical

• ex: intact classes, schools, or clinics

groups may differ from the start so causal claims are weaker; pretests, matching and statistical control help

Different Between Experimental, Quasi Experimental, Non-Experimental

What is Random Sampling

tells us who gets into the study

everyone in the population has an equal change of being selected

supports generalizing to the population

When is random sampling done and why

before participant completes the task

to make us confident that the study is generlizable

Random Assignment

who gets which condition/ decide which task participant completes

every participant has an equal chance of ending up in each group

makes groups comparable; supports causal claims

Relationship between Random Sampling and Random Assignment

you can have one without the other

can be an imbalance between both

ex: volunteers (not random sample) who are randomly split into 2 groups

When is NON-experimental design a good choice (4)

you are interested in a single variable

• how high are vowels in dialect A

You want a non-causal relationship between two variables if two things are related

• is there a correlations between speech rate and pitch

You want a causal relationship but you cant manipulate the IV or randomly assign participants (ethical or practical reasons)

• does damage to broca’s area reduce speech production ability

your question is exploratory

• which sounds are typologically most common across the worlds languages

Types of non-experimental design (3)

Cross-sectional

Correlational

Obervational

Cross Sectional

compare two or more “pre existing” groups (you don’t create the groups”

ex: whether speech differs with young vs old speakers

Correlational

see whether two variables are correlated

ex: speech rate and pitch

(more prevalent when both are numeric)

Obsevational

observe behavior in a natural or lab setting without manipulating anything

• see if 2 things are different

ex: record casual conversations and count discourse markers

Ethnography

become apart of the community and study their language

different from a hypothesis

• hypothesis is make on an ongoing basis

ex: go to a community in hopes on discovering something new

Control vs Generalizability

its a trade off

more control → stronger causal claims ut more artificial situation

combining methods (triangulation) helps ex: a lab experiment plus corpus data

internal validity

need to be confident the differences in the design is actually affected by the IV

can we be confident that the IV and not chance or extraneous variables produced the effect?

external validity

do the results generalize beyond this samble and setting

ex: from the lab to everyday speech

Triangulation

combining methods

List the Variables (2)

independent

dependent

Independent Variable

what you manipulate (experiment) or compare (non experimental)

• you dont measure you manipulate/compare against

also: factor, predictor, treatment, or explanatory variables

the groups within the IV are its levels (ex: age groups)

ex: speech rate (2 levels: slow, fast)

Dependent Variable

what you measure

what you are measuring

also: response, measured, or outcome variable (frequencey/count)

ex: rating of speaker’s persuasiveness

Extraneous Variables

anything other than my defined variable = extraneous

any variable other than the IV and DV

they are not the problem themselves as long as we CONTROL them

Confounding variables

also is known/suspected to have an effect on your IV

an extranous variable that is left uncontrolled and varies systematically withthe IV so its effect can’t be seperated from the IV’s effect

Lurking Variables

a (usually unmeasured) variable that influences both the IV and DV, creating a misleading association between them

Handing Extraneous Variables (5)

Hold constant

Randomzie

Counter Balance

Match or Block

Measure and include

Hold constant

same room, equipment and instructions for everyone

ex: 2 rooms for tracking exams one is very hot one is temperate → there will be effects from the room so put them all in the same one

Randomize

the best

random assignment spreads unkown differences evenly across groups co

counter balance

vary the order of conditions across participants (useful in within-subject designs)

order things differently

ex: 2 tasks to complete by the same person like fast and slow order the effect ex: 20 do fast first and 20 do slow first

match or block

pair participants on a relevant characteristic and split each pair across groups

define variables then try and block things

ex: IQ, define IQ then match 2 people with the same IQ they both go against each other

Measure and Include

record age, proficiency etc then add to the analysis

ex: age/proficiency they can’t control

collect background infothen include in the model so they are accounted for

Single-Factor Design

one independent variable (factor)

Univariate Design

one dependent variable

single-factor univariate design

one IV and one DV

Explain a within-subject design

each participant is exposed to all levels of the IV

Pros and Cons of Within-subject design

Pro: fewer pariticpants each person is their own baseline

Con: carryover effects (practice, fatigue, boredom, can figure out what is going on)

Between Subject Design

each participant is exposed to only one level of the IV

Pros and Cons of Between Subject

Pro: no carry over effect

Con: needs more participants; induvidial differences between groups

Within vs Between

counterbalancing in a within subject design

vary the order of the conditions across participants

Pretest-post test design

the classic controlled experiement

give a pretest to both groups

experimental group tgets the treatment; control group get “standard” material (a plecebo in clinical trials”

give a posttest to both groups

compare the change from pretest to posttest across groups

• to establish cause and effet

• establish the baseline for groups

• pretest → interrvention → posttest

Factoral Design

has two or more IVs (factors)

crossed: every level of one factor is combined with every level of the other(s)

lets us study several IVs simultaneously in a single experiment

What does factorial design reveal

Main effect: the overall effect of one factor (averaging over the other)

Interaction: the effect of one factor depends on the level of another

notation: a “2×3” design has 2 factors, the first with 2 levels and the second with 3 → 6 conditions

Mixed design

a single experiment with both a within-subject factor and a between subject factor

ex: every listener hears slow and fast speech (within) and listeners are either L1 or L2 speakers (between)

not to be confused with mixed methods

Mixed Methods

combining quantitative and qualitative strands in one study

Checklist: Reading any Study’s design