Research Methods in Speech-Language Pathology

1
Research Methods in Speech-Language Pathology
2
The Research Process An Overview

• Research is
• a systematic investigation designed to develop or
  contribute to generalizable knowledge
• a planned investigation of gathering and
  analyzing data from a sample to draw conclusions
  about a research question that can be applied to
  a larger population.
• From a discipline perspective, research is
• like diagnosis, particularly differential
  diagnosis, a processes of asking and answering
  questions
• a process of formulating answerable questions
  or forming testable hypotheses, and then making
  the observations needed to answer them.

3
The Research Plan

• Most research plans evolve from the following
  factors
• Observation of a problem that leads to a
  question
• Development of a problem statement in the form of
  a testable hypothesis
• Use of appropriate methods for testing the
  hypothesis
• Statement and interpretation of results and
• Discussion and evaluation of the results.
• Guiding the research plan are the rules of the
  scientific method that
• recognize that a problem can be studied
  objectively
• data can be collected through observation or
  experiment and
• conclusions can be drawn based on an analysis of
  the data that have been gathered.

4
Clinical Research and Evidence

• Like the research plan, the scientific method can
  be used to answer clinical questions and gather
  clinical evidence as well.
• The challenge for all practitioners is to answer
  the question How do you know that what you do
  works?
• Answering clinically relevant questions or
  testing clinically relevant hypotheses for
  clinical purposes should be no different than for
  research purposes.
• ASHAs Code of Ethics (2003) states that
  individuals shall evaluate services rendered to
  determine effectiveness.

5
Clinical Research and Evidence

• In order for clinicians to function ethically,
  they must systematically evaluate the impact of
  the services they render.
• They must move away from decisions based on
  opinion, past practice, and past teaching towards
  clinical decision-making that is guided by
  science, research and evidence (Elliott, 2004).

6
Evidence-Based Practice

• Evidence-based practice is more than clinical
  problem-solving.
• It requires SLPs to integrate their individual
  clinical expertise with the best available
  evidence from systematic (scientific) research to
  demonstrate that what they do works.
• Like the scientific method, evidenced-based
  practice involves the following steps

7
Evidence-Based Practice

• Converting a clinical need into an answerable
  question
• Searching for and finding the best evidence to
  answer the question
• Critically evaluating the evidence you find for
  its validity, strength, and applicability to the
  individual client
• Applying the results of the search and appraisal
  to clinical practice and
• Evaluating or auditing your performance.

8
Empirical Research

• Empirical research is based on observed and
  measured phenomena.
• It involves questioning, observing,
  experimenting, and eliciting behavior to
• define relationships
• demonstrate cause and effect and
• spark our own minds to begin thinking of other
  possibilities to be tested and studied.
• Empirical research leads to knowledge from actual
  experience rather than from theory or belief.

9
Empirical Research Designs Purpose

• Empirical research can be classified by its
  purpose or by the type of knowledge that will be
  produced.
• The three levels of classification are
  exploratory, descriptive, and explanatory.
• Exploratory (preliminary) research is typically
  conducted in the field and/or in natural settings
  or in non-natural settings with the purpose of
  discovering phenomena and/or theory.
• Exploratory studies require no active
  manipulation or alteration of the research
  context/condition by the researcher.

10
Exploratory Research

• It is typically conducted early in a research
  problem to uncover basic viewpoints, perceptions,
  behaviors, attitudes.
• It can be quite informal, relying on secondary
  research, such as literature or data reviews, or
  qualitative approaches, such as informal
  discussions with clients, employees, management
  or competitors.
• More formal approaches would include in-depth
  interviews, focus groups, case studies or pilot
  studies.

11
Exploratory Research

• Liebow (1967) conducted exploratory research
  through extensive observation and interview
  techniques to obtain stories of mens lives on
  Tallys Corner, an urban street corner in a black
  neighborhood to explore the meanings of their
  experiences.
• Liebow wanted to illuminate the richness of the
  lives of these men as well as struggles that each
  man encountered when attempting to fill varied
  life roles.
• His purpose in conducting an exploratory study
  was to reveal theory, not to support or test
  existing theory.

12
Exploratory Research

• The results of exploratory research are not
  usually useful for decision-making by themselves.
• The results can not be generalized because they
  are not representative of the whole population
  being studied.
• But they can provide significant insight into the
  best research design, data collection method or
  subject selection procedure.

13
Descriptive/Ethnographic Research

• Descriptive/ethnographic research involves the
  study of human behavior through the description
  or reconstruction of events that are observed as
  they naturally occur.
• The ethnographer collects naturalistic data which
  is then arranged to render complete descriptions
  of activities and interactions.
• Ethnography is an appropriate methodology for
  studying the customs, social patterns,
  speech-language behaviors, and rule-governed
  interactions of a culture or group of
  individuals.

14
Descriptive/Ethnographic Research

• With descriptive research, data are merely
  described and information is explained in
  relationship to the individual studied.
• Categories of analysis are generated and the
  relationship between the categories are
  explained.
• No inferences to a larger group or population can
  be derived from the analysis.

15
Descriptive/Ethnographic Research

• A classic example of naturalistic research is
  embodied in the work of Piaget (1932).
• Piaget studied language development by primarily
  observing and recording children's questions,
  reflections, and conversations.
• From such qualitative methods, he categorized
  various stages of what he termed egocentric and
  sociocentric speech development that, in turn,
  stimulated much additional research.

16
Explanatory (Experimental) Research

• Explanatory or experimental research is designed
  to reveal causal relationships and to predict
  outcomes.
• Such studies are founded on an accepted
  theoretical framework of reference and seek to
  support theory through hypothesis testing and
  prediction.
• It essentially involves comparisons of groups or
  individuals and relies on measurement and
  statistical analysis of quantitative data.

17
Explanatory (Experimental) Research

• Lets say for example you wanted to know if a
  hospital-based group socialization program was
  effective in reducing depression in young adults
  with a spinal cord injury.
• You might structure the study so that these
  patients were randomly assigned to either an
  experimental group, in which the socialization
  program was conducted, or to a control group,
  which received individual counseling.
• You would use a standardized depression scale
  that fit your theoretical framework.

18
Explanatory (Experimental) Research

• All participants would be tested on the scale
  before their participation in the experimental
  and control group programs and after the
  conclusion of the programs.
• A statistical comparison of the changes in
  depression scores between the two groups would be
  used to determine the extent to which there is a
  causal link between the socialization program and
  lowered depression.

19
Empirical Research Designs Structure

• Empirical research designs can also be classified
  by the way in which data gathering and analysis
  is organized and structured.
• The first classification system is used to
  examine and analyze the structure of
  experimental-type designs.
• There are three levels experimental,
  quasi-experimental, and non-experimental.

20
Empirical Research Designs Structure

• The second classification system is based on the
  nature of the data collected and the type of
  analysis conducted.
• Using this scheme, studies are classified as
  either qualitative or quantitative in their
  structure.
• Those that use numerical data and statistical
  analysis are quantitative.
• Those that rely on narrative and categorical
  analysis are qualitative.
• Lets start with looking at the structure of
  experimental-type designs.

21
Experimental Designs

• Experimental research is the appropriate method
  for investigating cause-and-effect relationships
  among variables, such as the effects of treatment
  on speech or language behavior.
• True experimental designs have the following
  characteristics
• Subjects are randomly assigned to at least two or
  more groups
• Some type of active manipulation is performed
  and
• One group of subjects (experimental group) is
  then compared with another non-manipulated group
  (control group) (Maxwell Satake,1997).

22
Experimental Designs

• Typically, quantitative measurements and methods
  of data analysis are used to compare the
  experimental and control groups.
• The results are presented as quantities or
  numbers (e.g., statistics).
• Based on the results, the researcher is able to
  support or refute the original hypothesis.

23
Experimental Designs

• Additionally, because experimental designs
  control for extraneous variables, the research is
  able to generalize findings to the larger
  population from which the sample is drawn.
• Nonetheless, for many practical and ethical
  reasons, it is sometimes impossible for an
  investigator to assign subjects randomly to
  treatment groups or to indiscriminately apply a
  particular treatment to one group while
  withholding it from another.

24
Experimental Designs

• Moreover, it could be argued that withholding a
  treatment from a target population or
  administering an alternative treatment with
  unknown effect rather than one with established
  benefits is unfair or perhaps illegal.
• This is often the case in many clinical studies
  in which an insufficient number of appropriate
  subjects may preclude the use of randomization
  procedures.

25
Quasi-Experimental Designs

• Quasi-experimental research designs are generally
  selected when true experimentation is impractical
  or impossible to perform.
• Typically, subjects are assigned to groups on the
  basis of preexisting conditions or circumstances.
• Suppose you work in a hospital clinic where you
  treat many adult patients for hoarseness
  accompanied by vocal nodules.

26
Quasi-Experimental Designs

• Following the diagnosis, the availability of
  therapy is on a first-come, first-served basis
  so that many patients are on a waiting list for
  three months or more.
• Although the use of randomization procedures may
  not be possible, you still wish to draw some
  conclusions about the efficacy of your treatment
  program.

27
Quasi-Experimental Designs

• To estimate treatment effectiveness, a
  constructed control group must be established
  for comparison with a treated group of subjects.
• The two groups would be matched on a number of
  variables prior to treatment.
• Matching is done to control as many extraneous
  variables as possible so that any positive
  between-group differences can be attributed to
  your treatment.

28
Quasi-Experimental Designs

• These variables would possibly include factors
  such as the degree of hoarseness, size of the
  nodules, duration of illness, occupation, age,
  sex, alcohol/tobacco consumption, etc.
• Quasi-experimental methods often necessitate the
  use of more control procedures.

29
Nonexperimental Designs

• Nonexperimental designs are those in which the
  three criteria for true experimentation do not
  exist.
• In such research, there is no attempt to achieve
  randomization, nor is any purposeful effort made
  to manipulate the variables under study.
• These designs examine naturally occurring
  phenomena and the researcher sets out to test and
  describe a concept or construct or the
  relationships among constructs.
• The researcher does not manipulate the
  independent variable but instead examines it in
  relation to one or more variables for descriptive
  or predictive purposes.

30
Nonexperimental Designs

• Any manipulation of variables is done post hoc
  through statistical analysis.
• Because random selection, manipulation, and
  control are not present in these designs,
  investigators must use caution when make causal
  claims from the findings.
• In health and human services, the three most
  common types of nonexperimental research include
  surveys, passive observation, and ex post facto
  designs.

31
Survey

• Survey designs are used primarily to measure
  characteristics of a population and are typically
  conducted with large samples.
• There is relatively minimal expenditure and
  numerous variables can be measured.
• Perhaps the most well-known survey is the U.S.
  Census which seeks to develop a descriptive
  picture of the characteristics of the population
  of the U.S.
• Statistical manipulation can permit multiple uses
  of the data set.

32
Passive Observation

• Passive observation designs are used to examine
  phenomena as they naturally occur and to discern
  the relationship between two or more variables.
• These designs are often referred to as
  correlational designs.
• Passive observation can be as simple as examining
  the relationship between two variables, for
  example, height and weight.

33
Passive Observation

• They can be as complex as predicting scores on
  one or more variables from knowledge of scores on
  other variables.
• Like surveys, variables are not manipulated but
  are measured and then examined for patterns.

34
Ex Post Facto Studies

• Ex post facto designs are a type of passive
  observation literally performed after the fact.
• The phenomena of interest have already occurred
  and cannot be manipulated in any way.
• For example, an ex post facto design could be
  used to examine differing career patterns of male
  and female social workers after graduation.
• One could look at patterns of gender differences
  in salary, career opportunity, geographical
  preference, etc.

35
Quantitative Methods

• Quantitative research methods are rooted in the
  tradition of physical sciences.
• With quantitative research, the researcher
  explores relationships using numeric data.
• Statistical methods are used to analyze the data
  and results are summarized in impersonal,
  objective reports.

36
Qualitative Methods

• Qualitative research methods enable researchers
  to study social and cultural phenomena using
  textual, rather than, numerical data.
• Case study, observation, and ethnography are
  considered forms of qualitative research.
• Results are used to understand and explain social
  phenomena in interpretive reports that reflect
  the researchers constructions of the data.
• As such, results are not usually considered
  generalizable, but are often transferable in the
  sense that readers will form their own
  constructions of what is reported.

37
Empirical Research Design Time

• Time is an important factor in the design and
  execution of research along both the experimental
  and naturalistic continua.
• There are two basic design classifications by
  time retrospective and prospective research.
• Retrospective research examines phenomena after
  the phenomena have occurred.
• Prospective research searches for cause and
  effect relationships or examines change in the
  present as the event unfolds over time.

38
Retrospective Research

• Retrospective experimental-type designs include
  those that use chart extraction and are
  classified as passive observation or
  correlational.
• A retrospective approach is useful when a
  reasonable data set, such as medical records, is
  available, if the researcher has limited time and
  funding resources, and if random assignment is
  not appropriate.
• Although there is no manipulation of a phenomenon
  or random assignment or control, the purpose is
  to describe the occurrence of phenomenon by

39
Retrospective Research

• Examining the relationships among variables and
• Examining possible causative relationships.
• For example, Davidoff et al. (1991) conducted a
  retrospective study to determine the efficacy of
  rehabilitation programs to facility recovery
  after acute stroke.
• They reviewed the medical records of 139 acute
  stroke patients at admission, discharge, and
  one-year follow-up from hospitalization.

40
Retrospective Research

• They examined such information as neurologic
  status, functional status, and the quantity and
  reasons of outpatient physical and occupational
  therapy.
• Another retrospective strategy is the life
  history approach.
• This approach to qualitative research uses the
  interview or informant technique to gather
  information or reminiscences about personal
  experiences, feelings, and events.
• This retelling and recasting of life events,
  feelings, and experiences can be used as a basis
  from which to interpret the sense of self as
  people age.

41
Prospective Research

• There are two basic types of prospective designs
  cross-sectional and longitudinal.
• In a cross-sectional study, the researcher
  examines a phenomenon at one point in time.
• There is only one single time frame in which data
  are collected.
• Longitudinal studies, on the other hand, involve
  data collection over extended periods of time.

42
Cross-sectional Study

• Suppose you wanted to describe attitudes toward
  aging and how individuals of varying ages
  perceive the aging process.
• You could design a cross-sectional study that
  sampled individuals in different age groups by
  conducting a survey or interview at one point in
  time.
• The analysis would involve a comparison among the
  age groups on the dependent variables, attitudes
  toward aging.

43
Longitudinal Studies

• Longitudinal studies can be used to examine such
  phenomena as the long-term effects of health care
  programs or interventions, the natural course of
  human development and adaptation, the trajectory
  of illness, or the sequelae of various diagnoses.
• There are three types of longitudinal studies
  trend, cohort, and panel studies.
• Trend studies involve examining a general
  population over time to see changes or trends
  that emerge as a consequence of time.

44
Longitudinal Studies

• Suppose you wanted to see if there were any
  trends in the way in which different age groups
  perceived the aging process over time.
• You would interview individuals who fit into the
  age brackets you are interested at one point in
  time, as in the cross-sectional design.
• Five years later, let us say, you would interview
  other individuals who fit the same age brackets.
• In this way, you would be able to compare and
  contrast attitudes toward aging of persons 41 to
  50 years old in 2000 with person of the same age
  range in 2005.

45
Longitudinal Studies

• Cohort studies involve examining a specific group
  or one particular generational grouping as they
  change over time.
• In a study on attitudes toward aging, we would
  examine aging perceptions of one age group at one
  point in time, such as 2000.
• Then, at a later point in time, let us say, 2005,
  we would sample different individuals but those
  who are from the same cohort as in the first data
  collection effort.
• The individuals in the sample would be five years
  older.

46
Longitudinal Studies

• Although we would be sampling from the same
  cohort of individuals we would not be
  interviewing the same individuals at both time
  periods.
• Panel studies are another longitudinal design
  strategy.
• Similar to a cohort design, the same set of
  people are studied over time.
• For example, the subjects we interviewed in 2000
  who were 41 to 50 years of age would be the same
  individuals we would attempt to interview five
  years later in 2005.

47
Research Variables

• When conducting empirical research, the notion of
  what constitutes a variable is important to
  understand.
• Variables are concepts or constructs that change
  under different circumstances rather than remain
  constant.
• Variables dont have to be numerical or
  quantitative they can be categorical, like
  gender.
• In the behavioral sciences, examples of variables
  would include
• stimulus characteristics (e.g., tone, intensity
  or frequency)
• environmental conditions (background noise level)
  
• speech behavior (rate of speech or number of
  nonfluencies)

48
Research Variables

• language performance (MLU in a language sample)
• hearing ability (speech reception threshold)
• gender (male, female, intergender)
• Variables can be categorized as independent or
  dependent.
• The behavior that the experimenter controls is
  typically referred to as the independent
  variable.
• Independent variables can be thought of as
  conditions that cause changes in behavior (B).
• In other words, the independent variable is the
  antecedent (A) in the equation ABC.

49
Independent vs. Dependent Variables

• The behavior that the experimenter measures is
  referred to as the dependent variable.
• Dependent variables can be seen as the behavior
  that is changed.
• In other words, the dependent variable is the
  consequent (C).
• For example, delay auditory feedback (the
  independent variable) may cause a change in
  speech rate (the dependent variable).

50
Independent vs. Dependent Variables

• In experimental research, the examiner
  manipulates the independent variable (while
  holding other potential independent variables
  constant) to examine the effect the manipulation
  of the independent variable has on the dependent
  variable.
• For example, MLU is sometimes used instead of
  chronological age to classify children into
  groups that vary in degree of language
  development.
• In this case, MLU is the independent variable.

51
Independent vs. Dependent Variables

• If, however, a researcher is looking at the
  effect of manipulating the number of
  communicative partners on childrens MLU, MLU
  becomes the dependent variable.
• The distinction between independent and dependent
  variables is really a distinction based on the
  use of variables rather than some inherent
  property of a variable.

52
Active vs. Attribute Variables

• A variable that can be manipulated is an active
  variable.
• The independent variable is an active variable
  when it can be manipulated in some way by the
  experimenter to see what effect it has on a
  dependent variable.
• For example, an experimenter can change the
  intensity of a tone presented to a listener by
  manipulating the hearing-level dial on an
  audiometer.

53
Active vs. Attribute Variables

• A variable that cannot be manipulated by a
  researcher is an attribute variable.
• Subject characteristics, such as age, gender,
  intelligence, type of speech disorder, degree of
  hearing loss, or history, have been manipulated
  by nature, but they cannot be changed by the
  experimenter.
• Some variables may be either active or attribute
  variables depending upon the circumstances of the
  research or on how the researcher uses the
  variable.

54
Active vs. Attribute Variables

• Anxiety is an example of a variable that can be
  active or attribute.
• Anxiety can be an attribute of subjects.
• Anxiety can also be manipulated by the
  experimenter by varying degrees to see what
  effect the manipulation of anxiety has on some
  dependent variable.

55
Continuous and Categorical Variables

• A continuous variable is one that may be measured
  along some continuum or dimension that reflects
  at least the rank ordering of variable values.
• The intensity of a tone, for example, can be
  measured along a numerical continuum from low to
  high values of sound pressure level.
• Stuttering frequency can vary from zero
  nonfluencies to a high number of nonfluencies.

56
Continuous and Categorical Variables

• A categorical variable cannot be measured and can
  only be categorized or named.
• Tones could be presented to a listener binaurally
  or monaurally.
• Subjects could be classified as stutterers or
  nonstutterers.
• Continuous and categorical variables are
  displayed differently graphically.

57
Continuous and Categorical Variables

• Change in a dependent variable as a function of
  changes in a continuous variable are commonly
  displayed in a line graph.

58
Continuous and Categorical Variables

• Change in a dependent variable as a function of
  change in a categorical variable is customarily
  displayed in a bar graph.