Research Exam 2

- Group designs

design

Formulate a plan, execute in a highly skilled manner

Research Design plan includes:

1. protocols for selecting participants

2. controls for extraneous variables

3. Observation of variables

4. Ensuring ethical procedures

5. implementation of tasks/treatments

Hypothesis is ——- AND ———

a prerequisite to develop adequate research design AND a tentative explanation for observation or clinical problem

the quality of a research design is judged by the ability to

answer research questions and control extraneous variables

Appropriate designs need to be able to answer

experimental questions or hypotheses.

a high level of internal validity ensures relationship between

independent and dependent variables

all possible individuals who have at least one characteristic in common

Target population or population of interest

extracted from target population with goal of making small observations for valid generalizations

sample

Goal for sampling:

1. extract a relatively small number of observations from population 2. this sample will serve as basis for valid generalizations about population as a whole

name the three sampling methods

simple random, convenience sampling, stratified sampling

Simple random sampling

individuals selected by chance process

substitute for simple random sampling, selected by practical constraints or geographical proximity

Convenience sampling

Sometimes this severely limits researchers’ ability to generalize results to larger population

convenience sampling

Stratified sampling

divide target population into a number of non-overlapping subpopulations; then random sample from within subpopulations

validity of study affected by

characteristics of participants

Previous research used to aid determine selection criteria of participants

authoritative resources

Include individuals from each constituency in the target population (women, minorities, etc.) as appropriate to the study

Representative samples

sample size is

the number of participants in the study

What contributes to the power of research design?

sample size or number of participants

What is statistical power?

The ability to detect a significant finding when one is truly present.

Power is affected by several factors:

internal validity, measurement reliability, choice of statistical tests, sample size

more participants =

more power to identify effects accurately (whether you have significant findings)

factors to consider in selection procedures

age, severity of disorder/impairment, education, handedness, etiology (if appropriate), time

post-onset or since onset of diagnosis (if appropriate), gender, lateralization, localization of brain-damage, vision, intelligence (if appropriate).

single group research design refers to

Observing one group of participants in two or more conditions

this research design is considered weak becasue it lacks scientific comparability

single group research design

in single group design, with no comparison group, you may lack

control of extraneous variables

examples of single group designs

1. pre-test - 2. post-test design: pre-test observation, experimental treatment, post- test observation

participants are compared to themselves in

single group design

uncontrolled nuisance.extraneous variables in single group research designs may be:

history factors, maturation factors, multiple-test effects, statistical regression effects, test sensitization or test practice effects

most serious problem with single group designs is that they

provide no means for knowing if extraneous variables have affected the DV

two group research design

Observation of two groups at different levels of IV

Time related nuisance variables (history and maturation effects) need to be controlled in

two group design

Participants are assigned to one of the two groups 2, One group receives treatment while other group serves as control; possibly receiving some type of stimulation but not an actual “treatment”

parallel design

More common research design in speech, language, and hearing

parallel

Crossover

participants alternate between treatment and control conditions

Each participant acts as their own controlMore economical because they require smaller number of participants 2. Problem with carryover from treatment to no treatment condition

crossover designs

require random assignment of participants to experimental and control groups

Independent research designs

how does randomization happen during independent design

flipping a coin, using list of random numbers, etc., ensuring comparability between groups

Related research design

quasi-experimental without randomization; random assignment is not possible because the IV is the classification variable and cannot be manipulated.

How does matching strengthen a research design?

Matching participants on several variables makes groups more equivalent.

importance of randomization

1. Randomization adds validity to statistical tests 2. Randomization minimizes confounding variables 3. Blinding can also reduce biases

Simple randomization

participants assigned to groups one at a time based on single sequence of random

assignments

Blinding reduces

the risk that participants’ expectations will influence the study outcome.

Goals of simple randomization

1) balanced in regard to individual characteristics 2) free of experimenter selection

biases

What is allocation concealment? what does it reduce? what does it improve?

1.Hiding the assignment of participants to groups from investigators. 2. reduces overstatement of treatment effects. 3. chances for reporting more accurate outcomes

for larger treatment studies or randomized clinical trials it is especially useful to have

allocation concealment

What is a benefit of hiding group assignments?

reduces bias when investigators know participant assignments.

how is group equivalence achieved in independent design

random assignment

In independent designs, participants are usually randomly assigned to groups to achieve

group equivalence

in related/quasi-experimental research designs, need to utilize ————— to achieve equivalence between groups

matching procedures

To achieve equivalence in related (quasi-experimental) designs:

1) identify relevant variables (those related to IV) 2) match two or more groups of participants on basis of these variables Ex: aphasic adults vs typically developing, match on age

issues for two group research designs: When is external validity weakened in pretest-posttest designs?

If the pretest sensitizes participants.

When should pretest-posttest designs be avoided?

When test procedures are likely to affect performance on a subsequent administration of a test., use a pre-test/post-test with different versions of the test.

Factors that can increase design complexity:

adding conditions to IV, adding IVs, more groups, combination

any design having more than 2 conditions of IV

Multivalent designs:

all possible combinations of levels of 2 more IVs 2. Effects of 2 or more IVs are observed within single study

Factorial Designs

How study variables jointly affect each other; the observed effect of one IV changes as a function of a second IV.

interaction of effects

What are the 3 types of designs? in complex research designs

RIM 1.Related: each participant experiences all treatment conditions. 2. Independent: matched groups experience a single treatment condition. 3. Mixed: combination of related and independent variables.

What can factorial designs include?

Additional levels for each variable and additional IVs.

multiple group designs include

two or more comparison groups along with experimental or treatment group

What does each comparison group typically serve as?

control for specific nuisance variables.

Adding a third control group that may not receive treatment.

Solomon four-group design

What are the two comparison or control groups used with language-impaired children?

language-matched and age-matched control group

Why use two comparison or control groups?

may yield more meaningful results than using just one comparison group.

variables in communication disorders are ——— distributed. Instead of ——

continuously, discrete

Distributions for categorical variables are shown in

bar graphs

continuous variables are usually displayed in

line graphs

theoretical distribution providing model for evaluation of variables

normal distribution

normal distribution is theoretical and has 3 characteristics

1) it is unimodal (one mode or “hump” at center) and symmetrical 2) it is continuous 3) it is asymptotic: curved line of the line graph  gets closer to horizontal axis as it moves away from the center

what is unimodal, continuous and asymptotic?

normal distribution such as weight, age, speech disorders

in normal distribution, curved line of the line graph  gets closer to horizontal axis as it moves away from the center is

asymptotic

all normal distributions are transformed to fit the

standard normal distribution

a pattern of scores that provides information about individual cases and the group as a whole.

distribution

What three major characteristics define a normal distribution?

unimodal and symmetrical, continuous, and asymptotic to the horizontal axis.

Normal distribution is ——— if It has a single peak or mode located directly at its center.

unimodal

Distribution curve line gets progressively closer to the horizontal axis as it moves away from the center without ever touching it, it is ———

asymptotic

What is the mean of the standard normal distribution?

0

What is the standard deviation of the standard normal distribution?

1

The standard score that measures how many standard deviations an observation lies above or below the mean is called a

Z score

What percentage of scores falls within ±1 standard deviation of the mean in a standard normal distribution?

68% of total scores

What percentage of scores falls within ±2 standard deviations of the mean in a standard normal distribution?

95%

What percentage of scores falls within ±3 standard deviations of the mean in a standard normal distribution?

99%

Why are proportion areas under the standard normal curve practically useful?

rx can determine exact probability of obtaining a specific outcome.

Z Score

standard score measuring individual location

What primary advantage does converting raw scores to Z scores provide?

direct comparison bw scores from different variables measured on different scales.

two issues with raw Z scores in clinical reporting?

include negative values and decimals, which can be misleading and difficult to interpret/communicate.

Bc Z score can be difficult to interpret and report, it is transformed into

distribution of standard scores with mean of 100 and SD of 15

skewed distributions are

not symmetrical and not bell shaped

skewed distributions: more scores with larger values toward right tail

negatively skewed

skewed distributions: more scores with larger values toward left tail

positively skewed

stat measure of the peakedness and tail thickness of a symmetrical distribution relative to a normal distribution.

Kurtosis

What does a sampling distribution of a statistic represent?

shows how often different values of a sample statistic occur when samples of the same size are repeatedly drawn from a population.

How does the mean of the distribution of sample means relate to the population mean?

both should be the same mean.

variability of the sampling distribution of means vs population variability

distribution of sample means is less variable than individual scores in the target population.

Standard Error of the Mean (SEM)

The standard deviation of the sampling distribution of sample means.

chance occurrence of particular outcome if sample of same size is collected from same population

probability distributions

do we usually know distribution of sample means?

no

standard deviation of sample means, how much a samples avg. is likely to fluctuate from true population average

Standard error of the mean (SEM)

Central Limit Theorem: Normality, what sample size is sufficient or normal?

greater than 30 (n>30)

estimate population parameters from sample parameters

parameter estimation