CSAD 148: Terms for Exam 1

Lecture 1: #1-8, Lecture 2: #9-26, Lecture 3: #27-36, Lecture 4: #37-41, Lecture 5: #42-56

Theory

a set of concepts and relationships among those concepts that helps us understand or explain some phenomenon

Evidence

a set of observations in the real world

* *develops methods observed to work in a sample*

Practice

used when an intervention works for a client

Evidence Based Practice

the process of applying a scientifically sound intervention technique to a particular client

* Three components:
* External scientific evidence
* Clinical expertise/expert opinion
* is this appropriate for this client?
* Client/caregiver perspectives
* client/caregiver should be given options if possible, but have overall autonomy over decision

Induction

type of reasoning that makes broad generalizations from specific observations

* there is data, then conclusions are drawn from the data
* from observations, clinicians/scientists discern a pattern, make a generalization, and __infer__ an explanation or a theory

Deduction

type of reasoning that starts out with a general statement and examines the possibilities to reach a specific, logical conclusion

* the __scientific method__ uses this to test a theory, idea, or relationship, form a hypothesis, and test the hypothesis in a set of observations

Inferencing

“carry forward” from data to theory

PICOT Questions

used to help find the evidence

* __P__opulation: who is the intervention effective for?
* __I__ntervention: what is the intervention?
* __C__omparison: what is being compared?
* __O__utcome: what do we want to change?
* __T__ime: how long does it take?

The Scientific Method

* a search for knowledge concerning general truths
* a process of __objectively__ studying a problem via collection of data
* a process of constant questioning
* a self-correction mechanism for knowledge acquisition
* a highly controlled version of clinical decision-making

Process:

* write a research question with measurable independent and dependent variables
* __test__ your hypothesis (plan and gather data)
* identify potential problems in your study
* write a new research question…

Database

organized portal which you can search multiple journals at the same time

* ex: Google Scholar, PsychInfo, EBSCO

Journal

a magazine-like publication of a scientific society

* ex: ASHA’s four journals; Journal of Speech, Language, and Hearing Research, American Journal of Speech Language Pathology, American Journal of Audiology, and Language, Speech, and Hearing in the Schools

Peer Reviewed Journal

a journal where the articles are only published after they have been reviewed by other researchers with content expertise

* often authors are required to revise and resubmit their articles
* only articles with strong research designs should be published in a peer reviewed journal

Peer Review Process

* when authors submit to a journal, the editor determines if the article might be a good fit


* if so, article is sent out to volunteer reviewers who write up a review of the article
* they can do the following; accept, accept with minor revisions, revise and resubmit, or reject

Double-Blind Peer Review

when neither the authors know the reviewer nor to the reviewers know the author

* this is the best quality, but can be meaner

Primary Study

studies that collect their own data

Secondary Studies

studies that review other people’s data

Systematic Review

* Plan for review is determined in advance
* Includes ALL studies about a given topic that you can find by searching multiple databases
* Includes a search strategy
* Narrows down to a smaller number of articles based on inclusionary and exclusionary criteria
* Reviews all the identified articles - no picking/choosing ones that match hypothesis
* Determines the level of evidence (quality of evidence) for each article
* Formulates an opinion based on ALL the evidence, focused on highest levels of evidence

Meta-Analysis

does everything a systematic review does, plus __re-analyzes__ the data from all participants in the previous studies using effect sizes

Randomized Controlled Trials

an experimental design in which the researcher adds or removes a variable to look at the impact of that variable of interest

* participants are randomly assigned to groups
* measured across groups
* measured across time
* gold standard of research

also:

* between subjects (treatment vs. control, or treatment A vs. treatment B)
* within subjects (DV is measured both before and after we manipulate the IV; pre vs. post)
* also has __random assignment__: participants are randomly assigned to treatment or control
* this attempts to eliminate the chance of group A looking different from group B *before* the treatment begins

Skeptical

open to new information, maintaining an attitude of constant questioning

Experimental Studies

implement a treatment to determine if the treatment was effective (like RCTs)

* without randomization, these studies are biased by subject selection
* Include an IV and DV

Observational Studies

simply watch (and compare or correlate) what is happening out on the world, do not implement treatment so you cannot say whether or not a treatment was effective

* lack *control*

Narrative Reviews

NOT systematic

* includes expert opinion including textbooks
* very bottom of study pyramid

Independent Variable

presumed __cause__

* the variable we manipulate
* determines if the question is comparative or correlational

Dependent Variable

presumed __effect__

* variable we measure the change of
* has to be the same or higher level of measurement as the independent variable

Causation

to have this, we must have

* relationship between IV and DV (correlation)
* temporal precedence (IV __before__ DV)
* NOTHING ELSE CAN BE THE CAUSE

Variable

anything that can change or be changed

* any factor that can be manipulated, controlled for, or measured in an experiment

Value

measurement of the variable

* metric of measurement

Nominal Values

aka name values, may be sets of 2 or 3

* yes/no
* yes/no/maybe
* rural/suburban/urban
* we code these with numbers, but they are not technically numeric

Ordinal Values

values that have a greater than/lesser than scale

* strongly agree/agree/neutral/disagree/strongly disagree
* less than high school/high school education/some college/college degree/advanced degree
* we will also code these as numbers, typically in order

Interval Values

go in order like ordinal values but with equal spacing

* proportional values
* these values are typically already in numbers
* temperature (fahrenheit or celsius), time (in minutes, hours)

Ratio Values

values that go in order (like ordinal) and with equal spacing (like interval) but also have a true zero and cannot be negative

* age (in years, months)
* height (meters, feet) or weight (lbs, kg)

Descriptive Questions

what is? (one variable)

* how many students in this class are first generation college students (percentage)
* what is the average age of students in this major (mean)

Inferential Questions

how are two (or more) variables connected?

* Do first generation college students take longer to graduate than students who have family members who have previously graduated college?
* Are students in CSAD, on average, younger than students in construction management?

Null Hypothesis

no difference/no correlation

* the treatment groups does not differ from the control group

Directional Hypothesis

specifies a value difference

* the treatment group __differs__ from the control group (non-directional)
* the treatment group was __more__ successful than the control group (directional; comparison)
* the number of hours of therapy was __positively__ associated with outcome (directional; correlation)

Quantitative Research

research involving numbers

Qualitative Research

research involving categorizations that we can code into numbers

Confounding Variable

other variables that may change the DV but we are not manipulating

* managed by having a control

Control

* done via inclusionary and exclusionary criteria
* also done mathematically (“after controlling for age, we found that X was associated with Y)

Double-Blind Study

one group of patients may be given a placebo, while another is given the actual medication that’s being tested

* neither patients nor the researchers know which group is taking the placebo
* the effects of the medication and placebos are then studied and compared with one another to determine the effectiveness of the medication itself

Observational Measurement

researchers observe the potential effect of a risk factor (eg. demographic variable), diagnostic test, or treatment strategy __without trying to manipulate__

* still have an IV (we just aren’t manipulating it) and a DV (outcome measure) and DV (outcome measure) and still use inferential statistics
* this type of research is important because we cannot randomly assign things like smoking, which why we cannot __prove__ the impact of smoking

Comparative Studies

studies that examine the difference in the DV between two or more different groups (IV)

* different from experimental studies because you cannot manipulate the IV
* use inclusionary and exclusionary criteria

Case-Control Study

researchers identify people with an existing health problem (cases) and a similar group without the problem (controls) and then compare them on a task

* type of comparative study

Retrospective Study

looking back, they already have the problem/concern/diagnosis when you begin the study

* type of comparative study

Cohort Study

researchers follow a cohort and track exposures to a variable of interest

* __compare__ the outcomes of the cohort on the basis of events/exposures
* __prospective__ looking forward
* type of comparative study

Cross-Sectional Studies

compare across groups

* cheaper
* faster
* more room for outside factors to have an influence (time)
* type of comparative study

Longitudinal Studies

compare across time

* within a group
* can __also__ compare between groups at multiple time points
* we do not manipulate the IV
* type of comparative study

Correlation

a __relationship__ or __association__ between two variables (IV and DV)

Correlational Studies

studies that identify relationships between variables of interest

* also may see the word “predict” to mean correlation
* they all run between -1 and 1

Bimodal Distribution

distribution with two modes

* if your correlation looks like a v-shape, you __cannot__ use correlational statistics

Participant Randomization

* reduces bias between groups
* no “opt-in” bias like in cohort designs
* but may not be ethical

Subject Matching

e.g.- age-matched, language-matched

* control the __participants__ by inclusionary and exclusionary criteria
* who is allowed to participate?

Counterbalancing Measurements

* control the __measurement__ order
* which test do you do better on, first exam or last?

Statistical Control

if you know you have a confounding variable, you can remove the correlation between CV and DV before examining the relationship between the IV and DV

Replication

repeating a study to see if you get the same results in a different sample (same population) at a different time

* have someone else do the study over
* exact/direct replication
* conceptual replication