Research: Purpose, Method, and Questions

Purpose of Research

• Research includes gathering information, making sound judgments, and developing new knowledge.
• The research cycle systematically progresses from a question, to a plan, to data collection/analysis, and finally information dissemination.

Scientific Method Steps

• The scientific method involves asking a question, followed by a background check.
• A hypothesis is then formulated and tested through experiments.
• The procedure is refined, if necessary, through repeated experimentation.
• Data is analyzed, conclusions drawn, and if the conclusions do not align with the initial hypothesis, a new research question is developed based on gained knowledge.
• Results are communicated.

Research Question

• A research question is a statement identifying the phenomenon to be studied.
• It focuses the study, determines the methodology, and guides all stages of inquiry, analysis, and reporting.

Characteristics of a Good Research Question

• A good research question is feasible, clear, significant (substantial and original), and ethical.

PICO: Creating a Research Question

• PICO helps to create a research question.
• Population considers patients/population.
• Intervention is the exposure to be considered.
• Control involves a control group (baseline, comparison treatment).
• Outcome looks at the outcome of interest and effect of intervention.

Research Paradigms

• Qualitative methods are wordy.
• Quantitative methods use numbers.

Types of Research

• Quantitative research confirms a hypothesis through numbers, with predetermined research design and variables.
• Qualitative research generates a hypothesis through gathered ideas, with an evolving research design and variable.

Approaches to Research: Qualitative vs. Quantitative

Worldview of Knowledge

• Qualitative research develops knowledge by examining patterns of meaning in people's words, actions, and records.
• Quantitative research develops knowledge through a collaboration of empiricism and rationalism.

Nature of Reality

• In qualitative research, reality is subjective and contextual.
• In quantitative research, reality is objective and can be known.

Approach

• Qualitative research uses inductive reasoning, moving from the specific to the general.
• Quantitative research uses deductive reasoning, moving from the general to the specific.

Steps in Reasoning

• Qualitative reasoning involves observing specific behaviors, identifying specific patterns, formulating a proposed hypothesis, and developing a conclusion, model, or theory.
• Quantitative reasoning starts with a theory, develops a specific hypothesis to test, collects observations to confirm or refute the hypothesis, and then confirms the theory.

Data

• Qualitative research uses ideas.
• Quantitative research uses numerical data.

Research Approach

• The research design and variables evolve in qualitative research.
• Research design and variables are predetermined in quantitative research.

Methods of Research

• Descriptive research involves observation to describe and understand patterns, including surveys, observational studies, epidemiological studies, and case studies.
• Predictive research identifies factors to predict relationships, often using correlation, but this does not imply causation.
• Experimental research identifies cause and effect factors through true, quasi, or pre-experimental designs.
  • Dependent variables are measured.
  • Independent variables are manipulated.

Variables

• Variables are entities that can take on different values.
• Independent variables are manipulated.
• Dependent variables are measured.

True Experimental Design

• Participants are randomly placed into groups.
• The manipulated variable is true.
• Eligibility is assessed, and participants are allocated to intervention, receive intervention, and are followed up for analysis.

Quasi-Experimental Design

• Causality is tested with suboptimal variable control.
• Participants are placed into groups non-randomly.
• Group subjects are based on a participant variable.

Experimental Research

• The control group represents the baseline.
• The experimental group may changed due to the independent variable.

Experimental Research (Cause and Effect)

• Pre-experimental designs have very little control group.
• Quasi-experimental designs often have a control group.
• True experimental designs always have a control group.
• Random selection of subjects from the population is absent in pre- and quasi-experimental designs, but present in true experimental designs.
• Random assignment of subjects to groups is absent in pre- and quasi-experimental designs, but present in true experimental designs.
• Random assignment of intervention to groups is absent in pre-experimental designs, sometimes present in quasi-experimental designs, and always present in true experimental designs.

Ethics

• Ethics are moral principles governing a person’s behavior or activity.
• The Code of Nuremberg (1949) is a ten-point statement outlining medical research on humans.
• Research is only justified if it is socially beneficial and conducted in a manner that satisfies moral, ethical, and legal concepts.
• The Declaration of Helsinki (1964) developed by the World Medical Association, is a main ethical statement in research articles today.
• The Declaration's main focus is on identifiable human data.

Research Ethics in Canada

• Researchers are responsible for the welfare of participants.
• Most journals require an ethical approval statement.
• Research Ethics Boards or Institutional Review Boards are required.
• Canada has three federal agencies: the Canadian Institute of Health Research, the Natural Sciences and Engineering Research Council of Canada, and the Social Sciences and Humanities Research Council.
• The Tri-Council Policy Statement: Ethical Conduct for Research Involving Humans, covers consent, conflicts of interest, data privacy/confidentiality, native status, and human biological/genetic research.

Core Principles

• Respect for persons includes respecting autonomy by providing the opportunity to deliberate about decisions
• Ensuring free, informed, and ongoing consent with free choice (voluntary), informed choice, and ongoing choice.
• Protect individuals with developing, impaired, or diminished autonomy with adequate information without coercion.
• Concern for welfare involves protecting a person’s quality of life in all aspects, including physical, mental, and spiritual health.
• Privacy and confidentiality must be ensured.
• Risks should be minimized and beneficence should be maximized.
• Accurate and accessible information should be provided.
• Justice includes fairness in treating people and equality in the distribution of research benefits and burdens.
• Deception involves lying about the true nature of a study because knowing the truth may alter behavior.
• Blinding uses a placebo.
• Debriefing is critical in studies with deception; researchers should provide information about the study's true purpose.
• Nutrition research in Indigenous communities has included nutrition experiments in residential schools, control vs. treatment groups of malnourished kids, and no consent.

Anatomy of a Research Paper

• The abstract provides an overview, including introduction/background, problem statement, methodology, findings, and conclusion.
• The introduction introduces the general topic, background, and previous research, and fills in the literature gap, as well as the purpose of the research statement, primary and secondary objectives.
• The literature review commonly is continued from the introduction, provides information on what has been investigated, can provide slight critiques, and rewrites the gap and why it’s important.
• The methodology outlines the research design, population, recruitment methods, measurement details, and analysis.
• The results section states relevant findings, demographics, completion and drop-out rate, outcomes specific to the study, without interpretation, and uses graphics.
• The discussion reviews findings, discusses outcomes, states the importance of the project, interprets findings, links the findings to literature, provides potential reasons for the outcomes, strengths, limitations, and future directions.
• The conclusion reiterates the study with main findings and importance, providing a concluding sentence about what should be done next and why without new data.
• Acknowledgements recognize contributions made by others who aren’t given authorship.
• Includes a list of references.
• Authors are listed by way of contribution.

CRAAP Test

• CRAAP (Currency, Relevance, Authority, Accuracy, Purpose) test is a method of determining the reliability of information included in the research.
• Currency indicates the timeliness of the information.
• Relevance refers to how relevant the information is in relation to the research.
• Authority refers to who the authors, sponsors, publishers, and sources are.
• Accuracy indicates the reliability, truthfulness, and correctness of content.
• Purpose describes the purpose or intent of the information.

Measurement Variables

• Focus: designing and conducting the study
• What data do we want to collect?
• How will we collect this data?

Research Variables

• Examples include height, weight, age, gender, scores on test, income, and birthplace.
• Concrete variables are easy to define, measure, and manipulate.
• Non-concrete variables have definitions and ways of measuring that are not obvious.
• All variables need to be operationally defined.

Operational Definitions

• Operational definitions specify the exact usage of a term in a study.
• They should reflect common usage in the discipline and are required for words that have multiple meanings.

Measurement Scales: NOIR (Nominal, Ordinal, Interval, Ratio)

• Non-parametric scales include nominal and ordinal.
  • Nominal scales indicate difference.
  • Ordinal scales indicate difference and direction of difference.
• Parametric scales include interval and ratio.
  • Interval scales indicate difference, direction of difference, and equality of units.
  • Ratio scales indicate difference, direction of difference, equality of units, and absolute zero.

Identity (Nominal)

• Each value has a unique meaning.

Magnitude (Ordinal)

• There is a specific order to the variables.

Equal Intervals (Interval)

• Data points along the scale are equal.

No Value of Zero (Ratio)

• Scale has a true zero where no data point acts as a starting point.
• Nominal = NAME (N & N)
• Ordinal = ORDER (O & O) + NAME
• Interval = INTERVALS (I & I) + NAME and ORDER
• Ratio = RADICAL (Think “Ratio is radical because it has it all”) + NAME + ORDER + INTERVAL

Parametric Stats

• Assumes data are normally distributed and follow a normal curve.
• Assumes data from all groups have the same spread.
• Assumes data are linear, with an example correlation.
• Often assumes data is independent.
• Linear correlation is on the graph.

Non-Parametric Stats

• Rely on few or no assumptions about the shape of the curve or parameters of the underlying population.
• Used in small sample sizes.
• Used with non-normal distributions.
• Not as affected by outliers.
• Generally, not as statistically significant.

Discrete Data

• Variables can assume only a finite number of real values within a given interval.
• Represents whole numbers that cannot be divided or broken down into fractions/decimals.

Continuous Data

• Variables can assume an infinite number of values.
• Can be broken down into units, fractions, and decimals.

Sampling Procedures: Population vs. Sample

• The population: all inclusive group of people that researchers generalize their findings to.
• Defined by researchers.
• A sample is a representative subset of the population that takes part in a study.

Representative Sample

• The sample must be like the population.
• Findings only relate to the group that was drawn.

Probability Sampling

• Involves random selection.
• In a simple random sample, odds are the same for everyone.
• Systematic sampling: researchers use lists to select every nth (or any number) person.
• Stratified random sampling: the population is divided into subgroups before randomization, ensuring subgroups are fairly represented, and then the population is ordered, followed by a random sample.

Non-Probability Sampling

• Population does not have an equal chance of being drawn.
• Findings may be generalized, but unlikely.
• Convenience sampling: researchers recruit participants that are convenient to them for their time and place; they are mainly volunteers.
• Quota: a certain number of participants are recruited.
• Generalizability: can the findings from a sample be generalized to the population or outside of the lab or research setting.
• Requires the reader to decide whether legitimate inferences or generalizations can be made.

Sample Size

• Equations and computer software exists to determine size of sample.
• Smaller sizes leads to type 2 errors.
• Goal is to have more data.

Random Assignment

• Also known as randomization.
• How the sample is put into the treatment groups.
• Important to prevent bias, and different treatment groups are equivalent.
• Typically computer generated.

Qualitive Sampling

• Generally, a sample is not taken from the population.
• Participants are selected for characteristics.

Purposeful Sampling

• Select sample from which the researcher can learn the most.
• Analyze and data collection simultaneously.

Quota Sampling

• Recruiting a certain number of participants with characteristics of interest, then finding people with the criteria until the quota is reached.

Snowball Sampling

• Using chain referrals.
• The participants refer the researcher to people they already have contact with who would satisfy the study's inclusion/exclusion criteria.
• Useful for groups that are not easily accessible via other strategies.

Data Collection

• Tools used in data collection.
  • Self-report: participants respond to the researcher's questions.
    • Measures how people think and report how they feel, act, think.
    • Can be quantitative or qualitative.

Question Delivery

• Self-administered: low cost, chance for misinterpretation, low response rate.
• Phone: higher response rate, higher cost, takes time, socially desirable response.
• Personal Interview: high response rate, easy clarification, bad for cost and time, interviewer bias.

Question Style

• Closed-ended: quantitative, participants must choose from a number of alternatives.
• Rating scales: quantitative, participants must provide a rating.
• Open ended questions: qualitative, asked to expand on answer.

Question Errors

• Loaded questions: contains unproved assumption or emotionally laden term- makes respondent feel guilty/ shameful.
• Leading questions: pushes an answer into the participants mouth and gives forced answers.
• Double barreled questions: asks multiple things at once.

Interviews

• Quantitative interviews are usually open-ended questions.
• Typically recorded and transcribed, unless not feasible or too intrusive.

Question Types

• Informal conversation interviews are unguided and most flexible, also hardest to summarized.
• Semi-structured interviews are semi-guided.
• Standard open-ended interview are completely structured, least flexible, easiest to summarize.

Test

• Is an instrument used to assess individual differences in various content areas.
• Possible problems- state of participant on given day and interpretation of results.

Behavioral Measures

• Observation and recording of behaviors.
• Possible problems- reactivity, observer accuracy.

Classifying the Observation

• Participant vs non-participant: participating or not.
• Disguised vs non-disguised- subjects aware or unaware.
• Laboratory vs natural setting- control or realism.
• Direct vs indirect- as it occurs/ active involvement or after the fact observation.

Case Study

• Involves a research question, setting, and a study of some kind.
• Measures of bodily activity, equipment, objectivity are also included.
• Possible problems: equipment accuracy, conductor skill.

Data Types

• Quantitative variables includes a numerical score, weight, age, VO2 max.
• Qualitative variables includes categorical variables that may represent a discrete category- hair color, shoe preference, dog breed.

Organizing/Summarizing Data

• Frequency distribution- information on number of observations for each variable.
  • All scores listed “f”= # of times each score occurs.
  • Show data pattern.
• With Histograms the bars do touch each other to indicate data is related.
  • The X-axis order is specific.
  • Good for Ordinal, Interval, and Ratio data.
  • Appropriate for discrete and continuous data.
• Frequency polygrams-can be expressed in frequency polygons. - Showcases data from cumulative frequency over time.
• Bar graphs: the bars do not touch each other.
  • The X-axis order is not specific.

Validity and Reliability

• Researchers are concerned with accuracy of measurement.
  • Accuracy of measurement- is it valid
  • Conisistancy- is it reliable
• Validity: the extent to which a test measures what it is supposed to measure
  • It is a trait of the instrument
  • Needs only to be established once
  • Proper use of the instrument should yield truthful results
• Types of Validity - Logical validity- weakest types and qualitatively: - Face validity: uses common sense approach, involving no statistics or numbers to express accuracy. - Content validity: use of written test/surveys, and a panel of experts that determine and assess logic and authoritativeness, with no values to express accuracy. - Statistical validity: - Criterion- based validity: - Concurrent: ability to estimate current performance, with used of a new measure to a gold standard, when 2 measures are obtained in close time proximity. - Predictive: Ability to predict future performance and the Event predicted is criterion measured and presented test is one to be evaluated.
  • Steps- -Develope an inventory of questions - Responses are score-based - Administer the questions to a representative sample who vary in smoking habits - Responce to the questions=score - Level of smoking= score - Statistically exime the degree of relationship between the two scores - Construct validity: Used when the variable has no definitive criterion, difficult to measure, can’t be directly observed
• Reliability - Error in results can happen due to the participant, testing, scoring, and instrumentation. - Correlation can be used to determine the degree of reliability. - Values closer to 1 have less error. - Measures the consistency of scores/data. - Informs about the dependability of the test results. - Refers to the score/data and not the instrument. - Must be established when reporting reliability values.
• Types of reliability. - External- comparing test - Internal- consistency of a test within itself

Measures of Central Tendency

• Mean- average of the data set: sum of data set/ number of data.
• Median- the middle of the data set: (n+1)/2.
• Mode- the most frequently occurring

Measures of Variation

• Range- the distance between the endpoints: highest score- lowest score.
• Standard deviation- indicates how far the values stray from the mean.
  • Square to remove any negative values.
  • Take square root to remove the effect of squaring.

Normal Curve

• Describes the overall shape of the distribution and range of a population mean,
• Includes a method for visually displaying data

Distribution Shape

• Kurtosis measures kurtosis and is identified by-
  • Shape of curve or tailedness.
  • How flat or peaked a normal curve appears.
• Named forms included-
  • Mesokurtic- Distribution resembles the normal curve
  • Platykurtic- Tails are short and thin/ Negative kurtosis because of lower peak
  • Leptokurtic- Longer and fatter tails/ Positive kurtosis

Skewed Distributions

• Involve many distributions that are not normally distributed.
  • These can present as scores that cluster, opposite the tail.
  • Or with possitively skewed distributions that has greater mean that median.
• Others can offer Negative distributions for which the mean is less than the median

Descriptive Statistics

• Measures of Central Tendency offer information about the middle of a data set.
• Measures of Variability offer provide information about how spread out the data is.
• Distribution shape provides information about whether the majority of the data clusters near the mean or spreads throughout the data set.

Z-Score

• Measures skewness as the measure of the lack of symmetry of a data set and indicates a non-normal data distribution.
• Can have negative- for the area under the curve from the mean.
• Calculations involve: Z score- a measure of how many units a raw score is from the mean.
  • z=x-μ/ σ: x is the raw score, μ is the mean, and σ is the SD.
  • Positive scores above μ, negative scores under, 0 means the mean.

Percentile Rank

• Score indicates % of those which are at or below the given raw score.

Hypothesis

• Hypothisis : An educational guess with limited evidence as a starting point.
• Hypothesis testing : The process of determining whether a hypothesis is supported by the results of research
  • A process that includes Accepting or rejecting study based on objective and logical stats and referring to the population.
• Null hypothesis: Is defined as a statement that ther is no difference.
  • Generally assumed to be true, until proven.
  • Aim to see if the null hypothesis is true and the purpose of the research is to decide whether Null hypothesis is probable true or probably false.
• Alternative hypothesis which Predicts that a significant difference exist between groups being compared for study. -Is Used when null is rejected. -One-tailed hypothesis: Includes research that predicts the direction of the expected difference between the groups.
• Two-tailed hypothesis: researcher expects to find a difference between the groups, but does not predict the direction of that difference.

Hypothesis Testing Terminology

• Ho: Default position, Generally assumed to be true until evidence indicates otherwise.
• Ho is True Groups Truly the Same: Result "true negative".
• Ho is False Groups Truly Different : Result "true positive".

Statistical Significance

• Defines an differences is groups is unlikely to have occurred by chance, generally 5% alpha level. -Common misunderstandings: that the probability of true or false is not statistically signigicant.
• Type one results are generally more serious. How to reduce both. -Reduce the alpha- level. -Increases the chance of making a type 2 error -Increase sample size to help.

Calculations in Testing and Power anlysis :

• P-values, power, alpha, beta.
  • P-value is an also known as an alpha level,
  • Power = Inverse of beta(Power = 1-β), defined a the Probability of what must be 11.1 in test review?

confidence Intervals:

• Confidence intervals(CIs) indicate the estimated range of values that is likely to include the unknown population parameter. -Population parameter: is defined as a Number that describes a fixed but unknown characteristic of the population. -Which indicates the Mean of your estimate +/- the variation of that estimate.
• Confidence that indicates probability
  • Interval =range you expect your that shows estimate to fall between What makes for Confidence interval success: -Instead of estimating the parameter by a single value, use an interval that is likely to include the parameter given -Which provides A confidence interval that provides a lower limit and an upper limit that is likely to contain the true population parameter

Measuring and Identifying "X" Factors

To measure:

• A 68% confidence interval to mean that there is a 68% chance that the calculated confidence interval contains the true population parameter
  • The Z factor is z* to use is 1.00
• A 90% means the confidence interval contains a 90% chance that the calculated confidence interval contains the true population parameter.
  • In this The Z factor is z* to use is 1.645
• A 95% indicates confidence is a 95% chance that the calculated confidence interval contains the true population parameter.
  • Here the z* to use is 1.96

Relationships Between Items"

A 99% that the confidence interval to mean contains has a a 99% chance that the calculation is related. -The Z is set to z* to use is 2.58 Reporting the findings;

• The conclusion must Express the confidence level in percent with the lower limit(LL) and upper limit(UL)
  • To represent 95% CI [LL, UL]
• Testing for Misconception: In the Event that When comparing two parameter estimates, -It is almost always true that the Cls do not overlap, then the result will be statistically significant or correct: -Is it not correct to assumes that if CIs do overlap, that the result is not statistically significantly

Establishing Correlation Between Items

• Describes degree to which variables are linearly related(positive or negative) and potentially be used for predict(predictive validity).

  • Must be Used when impractical or unethical to do an experimental study

• Test The null hypothesis that there is no relationship between the variables and find or establish the alternative hypothesis.

• Correlation coefficient: uses the Pearson correlation coefficient and measures Measures of the correlation between 2 variables (interval or ratio scale) Which has a Relationship that is expressed as r, and which offers a R value that is measured from -1 to 1 and allows for Interpertation: -1 is strong 0 is no relationship +1 is strong, the 1's are perfect.

• As "r" approaches As -1 or +1 there is an increases correlation between the two variables, -However you Can only tell us about linear relationships

Measuring Relationship Types

If r is found to be between 0 and 1 the Positive correlation is expressed in items changing in the same direction -Example: A puppy's length and weight; good customer service and sales; hours spent studying and exam grade; speed of wind turbine and energy produced If r= 0 there is a No correlation or no relationship, as seems in examples of, Amount of candy consumed and length of hair; Car speed and shoe size; running speed and toothpaste preference Between 0 and -1 this is a Negative correlation, which Suggests if one variable changes, the other will change in the opposite direction. An Example can be found in cases where Amount of exercise and blood pressure; a hen's age and number of eggs produced; caffeine consumed and perception of fatigue

Validity of strength in relationships:

• The more the no relationship, the weaker the relationship.
• As the relationship increases so to does the strength to the relationship: -0.761.00 = High -0.51-0.75=Fair -0.26-0.50=Moderate -0.00-0.25=None to weak * or curvilinear

When using Validity one must * Careful!!! and Graph results to make sure it is not curvilinear

• Scattergrams/Scatterplots provides diagrammatic representation of the relationship between 2 variables, -With what gives a quick indication of relationship. It is Important to graph-relationship in case a non-linear relationships exists.

###Relationship types Positive -Where Variables move in same direction . -A rise to and increase (decrease) in one produces an increase(decrease) in the other. Negative