Introduction to Quantitative Research Methods PDF

Introduction to Quantitative Research Methods Required Readings: Chapters 9, 10, 15 (Woo, 2018)  Key Research Design Features  Intervention  Comparisons  Control over confounding variables  Blinding  Time frames  Relat...

Introduction to Quantitative Research Methods Required Readings: Chapters 9, 10, 15 (Woo, 2018)  Key Research Design Features  Intervention  Comparisons  Control over confounding variables  Blinding  Time frames  Relative timing  Location (Woo, 2018)  Broad Design Options Quasi- experimental Experimental Nonexperimen (controlled (randomized tal trial control (observation without trial [RCT]) al study) randomizatio n) (Woo, 2018)  Table 9.2 Hierarchy of Designs for Different Cause-Probing Questions Type of Hierarchy of Designs Question Therapy RCT/Experimental > Quasi-experimental > Cohort > Case Control > Descriptive correlational Prognosi Cohort > Case control > Descriptive correlational s Etiology RCT/Experimental > Quasiexperimental > Cohort > Case control > / Descriptive correlational harm (prevent ion) (Woo, 2018, p 144)  Therapy & Etiology/harm Prognosis Questions (prevention) Questions RCT/Experimental Cohort Quasiexperimental Cohort Case control Case control Descriptive correlational Descriptive correlational (Woo, 2018, p 144) Experimental Designs  Experimental Design Offer strongest evidence of cause and effect High on evidence hierarchy Researcher does something to some subjects; introduces an intervention Researcher introduces control Experimenter assigns participants to intervention or control condition on a random basis (Woo, 2018)  Experimental Design Advantages—most powerful for detecting cause and effect relationships Disadvantages—often not feasible or ethical, Hawthorne effect (knowledge of being in a study may cause people to change their behavior), often expensive (Woo, 2018)  Variations Posttest-only (or Outcome data after-only) collected only after design the intervention Pretest–posttest Outcome data collected (before–after) at baseline and after design intervention Crossover Subjects exposed design to 2+ conditions in random order (Woo, 2018)  Intervention –vs- Control Conditions Intervention “Treatment” described in formal protocols Importance of intervention fidelity, or if treatment delivered and received as planned “Control” gets no treatment; no intervention “Usual care” or standard or normal procedures used to treat patients Placebo presumed to have no therapeutic value if used (Woo, 2018) Quasi-Experimental Designs (trails without randomization)  Quasiexperimental Design Nonequivalent control group Involve designs intervention, but lacks Two Categories  randomization or control Within-subjects designs (Woo, 2018) Quasiexperimental Design May be easier and more practical than true experiments More difficult to infer causality Usually there are several alternative rival hypotheses for results (Woo, 2018) Nonexperimental Designs (Observational Studies)  Observational Design Also known as: Non-experimental design Not all independent variables can be manipulated Ex: gender; smoking Disadvantages: Does not yield persuasive evidence for casual inferences Advantage Efficient way to collect large amounts of data when intervention/randomization is not possible (Woo, 2018)  Correlational Studies Cause-probing questions (e.g., prognosis or harm questions) for which manipulation is not possible are typically addressed with a correlational design Correlation is an association between variables and can be detected through statistical analysis Such studies are weaker than RCTs for cause- probing questions (Woo, 2018)  Prospective Studies (cohort design) A potential cause in Prospective designs the present (e.g., are stronger than experiencing vs. not retrospective designs experiencing a in supporting causal miscarriage) is inferences, but linked to a neither is as strong hypothesized later as experimental outcome (e.g., design depression 6 months later) (Woo, 2018)  Retrospective Studies An outcome in the One retrospective present (e.g., design is a case– depression) is linked control design in to a hypothesized which “cases” (e.g., cause occurring in those with lung the past (e.g., cancer) are compared having had a to “controls” (e.g., miscarriage) those without lung cancer) on prior potential causes (e.g., smoking habits) (Woo, 2018)  Descriptive Studies Some research is descriptive (e.g., ascertaining the prevalence of a health problem) Other research is descriptive correlational—the purpose is to Not all Purpose is to describe whether research is observe, variables are cause - describe, and related, without probing document aspects ascribing a cause- of situation and-effect connection (Woo, 2018) Time Dimension in Research Design Cross-sectional - data collected at a single point in time Longitudinal - data collected two or more times over an extended period (follow-up) Longitudinal designs better at showing patterns of change and clarifying whether cause occurred before effect Challenge in longitudinal studies is attrition or loss of participants over time (Woo, 2018) Techniques of Research Control controlling controlling the participant randomization study context factors statistical matching homogeneity control (Woo, 2018)  Statistical Validity Low statistical power (e.g., sample too Statistical conclusion small) validity - ability to detect true Threats  relationships Unreliable statistically implementatio n of a treatment-low intervention fidelity (Woo, 2018) Selection biases arising from preexisting differences between History threats are events co- groups; single biggest threat to occurring with causal factor studies that do not use that could affect outcome experimental design Threats to Validity Maturation are processes that Mortality/attrition threat is a occur with passage of time loss of participants (Woo, 2018) Sampling Concepts  Sampling Concepts Population (“P” in PICO questions) Entire group of interest based on eligibility criteria Sampling Selection of portion of population (a sample) to represent entire population Eligibility criteria Characteristics that define population (Inclusion/Exclusion criteria) (Woo, 2018)  Sampling Concepts Sampling bias: over- or under- representing population in terms of key characteristics Target population: entire population of interest Accessible population Portion of target population accessible to researcher from which sample is drawn (Woo, 2018) Sampling Goal in Quantitative Research Representative sample More easily achieved Sample whose key with: characteristics Probability sampling closely approximate Homogeneous those of population populations Larger samples achieved through power analysis (Woo, 2018)  Probability Sampling Involves random selection of elements: each element has an equal, independent chance of being selected Stratified Systematic Simple random: random sampling: sampling: sampling population is involves establish a divided into two selection of sampling frame—a or more strata, every _th case, list of from which such as every population elements 10th person on a elements randomly list selected (Woo, 2018)  Nonprobability Sampling Quota sampling: Convenience identify population Purposive sampling: select strata and sampling: most conveniently figure out how handpicking available people many people sample members as participants needed from each (Woo, 2018)  Sample Size Number of study participants in final sample Sample size adequacy is key determinant of sample quality in quantitative research Sample size needs can and should be estimated through power analysis Risk of "getting it wrong" (statistical conclusion validity) increases when samples are too small (Woo, 2018) Data Collection Methods  Structured Self-Reports Data collected with formal instrument Interview schedule : Questions pre- specified but asked orally (Face-to- Questionnaire: face or telephone) Questions pre- specified in written form, & self- administered by respondents (Woo, 2018) Structured Self-Reports Open-ended Close-ended questions question (Woo, 2018)  Structured Self-Reports Advantages of Questionnaires Advantages of Interviews less costly higher response rates Advantageous for geographically appropriate for more diverse dispersed samples audiences Offer possibility of anonymity some people cannot fill out a (which may be crucial in questionnaire obtaining information about certain opinions or traits) opportunity to clarify questions or to determine comprehension opportunity to collect (Woo, 2018)  Psychosocial Scales Scale- device that assigns numeric score along a continuum Likert scales- summated rating scale Visual analog scales (Woo, 2018)  Response Set Bias Biases reflecting tendency of some people to respond to items in characteristic ways, independently of item content Social desirability response set bias Examples Extreme response set Acquiescence response set (yeah- sayers) (Woo, 2018)  Observational Methods Structured observation of pre-specified behaviors Concealment Category systems  checklists (Woo, 2018)  Evaluation of Observation Risk of observational biases - factors that can interfere Excellent for Potential problem with objective observation capturing many of reactivity Observational biases probably clinical when people aware cannot be eliminated, but can phenomena and they are being be minimized through careful behaviors observed observer training and assessment (Woo, 2018)  Biophysiologic Measures In vivo measurements- Performed directly within or on living organisms In vitro measurements- Performed outside organism’s body (e.g., urinalysis) Accurate, objective, valid, and precise May be cost-effective for health researchers Caution may be required as advanced skills may be needed for interpretation (Woo, 2018)  Factors Affecting Data Quality Procedures used to collect data Circumstances under which data were gathered Adequacy of scales used Psychometric assessment evaluates measurement properties Reliability is the extent to which scores are free from measurement error (Woo, 2018)  Interpretation of Quantitative Results In addition to Statistical results statistical Results supporting do not communicate significance, hypotheses are much meaning and effects must be statistically must be interpreted considered if large significant for usefulness and clinically important Results interpreted for precision of estimates Important to (confidence intervals) & clinical decision magnitude of effects (effect making sizes) (Woo, 2018) Clinical Significance (Woo, 2018)  CONSORT Guidelines   (Woo, 2018)