NDFS 475 Research Methods and Practices Study Guide

· What is the definition of "research" from a scientific standpoint?

Systematic investigation to establish facts or principles.

Research Process

1.identify the research question

2. begin reviewing the literature

3. formulate a hypothesis

4. continue to review the literature and refine the question

5. develop a research approach

6. justify and fund the study

7. obtain approval (IRB), ethical considerations

8. execute the study

9. analyze and interpret the findings

10. formulate conclusions

11. communicate findings through peer review

12. incorporate significant and relevant findings into practice

look at people at a certain point in time

following same people over a certain period of time (can be longitudinal or correlational)

does intervening and implementing a new practice/supplement have an effect

cross sectional

cohort study

Intervention study

Practice-Based Research Cycle

Current practice → generate question → review literature → design and implement study → interpret data → determine relevance to practice → disseminate findings → apply new knowledge (cycles back through)

· What activities are important for professional development?

· How can quality journals/publications be evaluated?

· Be able to distinguish between and recognize research questions, hypotheses and purposes/objectives

Typical components of a hypothesis

Measurable

Specifies the population being studied

Specifies the type of relationship being examine

Defines the variableIdentifies a time frame

States the level of significance

Research questions are more broad than _________

specific research objectives

PICO/PICOT/PICOMT (in the research question)

Patient, population, or problem

Intervention

Comparison

Outcome

Type of question or methodology

Type of study (or time)

Hypothesis vs theory

H = A logical supposition/reasonable guess; an educated conjuncture to explain a relationship between two or more variables

Theory = An organized body of concepts, principles, and facts intended to explain a particular phenomenon. (much more developed than a hypothesis; you take a theory and test hypothesis against that theory)

· What is a null hypothesis vs an alternative hypothesis?

null = There is no difference between the variables. The most commonly used hypothesis for statistical analysis.H0; μ1 = μ2

Alternate = Opposite of a null hypothesisHa; μ1 ≠ μ2

· What are ordinal, nominal, interval, and ratio data?

Continuous scores that have a common unit of measure between each score and a true zero point (weight, length, speed, distance, calories)

Categorical scores that cannot be hierarchically ordered (gender, marital status, geographic area, ethnicity...)

Continuous scores that have a common unit of measurement between each score, but do not have a true zero point (temperature, time on a clock with hands, IQ or SAT scores)

Categorical scores that do not have a common unit of measurement between each score, but are ordered high to low (number of sit ups and then ranked; pain severity; likert)

Ordinal: Categorical scores that do not have a common unit of measurement between each score, but are ordered high to low (number of sit ups and then ranked; pain severity; likert)

Nominal Data = Categorical scores that cannot be hierarchically ordered (gender, marital status, geographic area, ethnicity...)

Interval data = Continuous scores that have a common unit of measurement between each score, but do not have a true zero point (temperature, time on a clock with hands, IQ or SAT scores)

Ratio data = Continuous scores that have a common unit of measure between each score and a true zero point (weight, length, speed, distance, calories) For example, zero weight means no weight at all.a

· What are dependent and independent variables?

Dependent = Outcome variables, the effect

Independent = held constant, controlled, known, cause, intervention

· Why is a literature review important?

Brings the researcher current on the topic

Forces reviewer to see what has been done

Helps refine the question

Brings to light appropriate methods for data collection and study design

Justifies the project

Helps you find out who the experts are in the field

Justifies national guidelines or evidence based practice statements

·Know the components and proper sequence of a traditional research report

Introduction (statement of the problem; purpose/objectives/aims; research questions; hypothesis)

Review of the literature

Methods

Results

Discussion/conclusion

References/appendices

· Understand and define the different categories of research

Introduction (statement of the problem; purpose/objectives/aims; research questions; hypothesis)

Review of the literature

Methods

Results

Discussion/conclusion

References/appendices

·Understand the different types of study designs (descriptive and experimental), and identify the basic strengths and weaknesses of each one

Descriptive Study Design

-Describe a population of interest

-Most common starting point and less money

Experimental

-To investigate cause-and-effect relationships by manipulating one or more variables.

Types of descriptive research

-Cross sectional

-surveillance

-case report

Repeated measurements of the instrument result in the same value within the same subject. There is good reproducibility/repeatability

Consistent answers every time. A method with high precision.

The instrument gives substantially the same value as the true value (the method is precise and has no bias). Usually reported as a percentage of the true error. [you may get the same lab value every time, but it may not be the correct value]

the instrument or test accurately measures what it is supposed to measure. It is reliable and has no bias (you measure what you thought you measured)

Accuracy

Precision

validity

Reliability

Precision

Reliability

Accuracy

validity

Measurement error. Inaccuracy due to random non systematic error (measurement error) or a systematic error (bias)

bias

the percentage of persons with a disease or condition who have a positive test result

'

the percentage of people without the disease or condition who have a negative test result

sensitivity

Specificity

sensitivity

Specificity

If false positives are unacceptable such as diagnosing celiac's or confirming a surgery-needing cancer, then ______ is more important

Specificity

The percentage of people with a positive test result who actually have the disease

Positive Predictive Value

Probability that subjects with a negative test truly do not have the condition.

Negative Predictive Value

Probability of a positive test result for a person with the disease divided by the probability of a positive test for a person without the disease

Likelihood Ratio (LR)

Receiver Operator Curves (ROC)

Probability of a positive test result for a person with the disease divided by the probability of a positive test for a person without the disease

I think its just like the likihood ratio

Voluntary consent created. Created 10 ethic statements.****

ethical code of conduct for research that uses human subjects

o Nuremberg Code

o Declaration of Helsinki

o National Research Act

o The Belmont Report

Nuremberg Code

created IRB

o Nuremberg Code

o Declaration of Helsinki

o National Research Act

o The Belmont Report

Declaration of Helsinkiz

creates national commission for protection of human subjects

Result of the Tuskegee Syphilis Study*

o Nuremberg Code

o Declaration of Helsinki

o National Research Act

o The Belmont Report

National Research Act

THREE BASIC PRINCIPLES

1. Respect for persons (autonomy);

2.Beneficence (maximize benefits and minimize harm);

3. Justice (burdens and benefits of research should be justly distributed

o Nuremberg Code

o Declaration of Helsinki

o National Research Act

o The Belmont Report

Belmont Report

· What is an Institutional Review Board (IRB) and what is its purpose?

Institutional Review Board protects the rights and welfare of human subjects involved in research activities; ensure 10 ethical directives are being followed

· What are the components of informed consent and when is it required?

6-8th grade reading level

Describes nature of study and what's involved (activities and duration)

Can withdraw at any time

Research contact info

Offer to provide results of study

List of potential risks or discomforts

Guarantee of confidentiality and anonymity

A signature line

· What are Type I and Type II errors in hypothesis testing?

Type I Error (False Positive): You reject the null hypothesis when it is actually true.

Type II Error (False Negative): You fail to reject the null hypothesis when it is actually false.

Type I Error: You think there's a wolf (reject the null), but there is no wolf (null was true).→ False alarm.

Type II Error: You don’t see a wolf (fail to reject the null), but a wolf is really there (null was false).→ Missed danger.

Collect data to follow trends on a population or subpopulation. Not the same cohort. Generally longitudinal. Data collected periodically. Can determine descriptive statistics (prevalence, incidence)

Population/surveillance studies

Strengths:

Identifies potential risk factors

Identifies the scope of a problem

Data is available to the public for future research

Weaknesses:Does not show causation

Snapshot of a group at one point in time. Can determine relationships, associations, correlations, provides descriptive statistics for a subpopulation

Cross-sectional studies

Strength: inexpensive and shows relationships

Weaknesses: Does not show causation

a type of retrospective study. Subjects who have condition or common characteristic are compared to a control group w/o the condition. Can determine relative risk, odds ratios; associations

Case control studies

Strengths:

Identifies risk factors

Outcomes of disease

Morbidity and mortality

Weaknesses:Does not show causation

Follow a group with a common interest, it is prospective and longitudinal but there is no intervention. Strongest type of observational study. Can test a causal relationship so has an analytical approach with a defined endpoint. Can determine incidence, association, relative risk, descriptive statistics, x^2, logistic regression [Framingham, Nurse's Health Study]

Cohort studies

Strengths:Longitudinal

Weaknesses:

-no controls

expensive

descriptive research to describe characteristics, attitudes, behavior, opinions, knowledge etc. in a certain population. Can determine relationships, correlations, descriptions

surveys

Strengths: Informative

Weaknesses:

Does not show causation

Self reported data

single or multiple observations but no formal study design or method to gather data

Case report

Strengths: generates a question or hypothesis

Weakness: doesn't prove anything

intervention is given to a single group, there is no control group. Suggests cause and effect.

Uncontrolled trials (quasi experiment)

Strengths: Less expensive than RCT's

Weaknesses: No control group, can't show causation

Intervention but not randomized, prospective. Suggests cause and effect; efficacy, magnitude of effect. Statistics include t-test, ANOVA, x^2, logistic regression, relative risk

Nonrandomized trials (quasi experiment)

Strengths: Less expensive than RCTs

Weaknesses:Not randomized so not able to extrapolate beyond study group

Control group, treatment group, intervention. Random assignment into groups, researchers manage the exposure to the treatment, prospective study, should be double blinded when possible. Can determine cause and effect, efficacy, or magnitude of an effect. Statistics include: t-test, ANOVA, x^2, logistic regression, relative risk

RCT

Strengths:provides clearest evidence of a causal effect

Weaknesses:ExpensiveComplex

Reject null hypothesis but it's true. Aka the - error (false positive)

Fail to reject the H0 (there is a difference but we didn't see it) so you accept the null hypothesis when there is a difference (false negative)

Typer 1 Error

.Aka the alpha error

- Reject H0 but it's true. Aka the alpha error

Typer II

-Fail to reject the H0 (there is a difference but we didn't see it)

S

predicting the number of subjects needed to detect an effect if there is one. Requires estimated acceptable variance (SD), desired alpha and beta level

power analysis

How do we reduce the risk of committing Type I and II errors?

Type I (alpha error) = decrease the alpha so .05 to .01

Type II (beta error) = increase the sample size

· How do we statistically evaluate measures of variance in data?

Range

Standard Deviation

Standard Error

Confidence intervals

Sampling Error

Probability that a population parameter will fall between two set values

error from having a sample rather than the whole population

average deviation of scores around the mean

range of scores from low to high

measurement of sampling error. the SD of sampling distribution. The larger the N, the smaller the SE. A larger SE means the less confidence one can be in the estimate representing the population.

Range = range of scores from low to high

Standard Deviation = average deviation of scores around the mean

sampling Error = error from having a sample rather than the whole population

Confidence intervals = Probability that a population parameter will fall between two set values

Standard Error = measurement of sampling error. the SD of sampling distribution. The larger the N, the smaller the SE. A larger SE means the less confidence one can be in the estimate representing the population.

· What is the difference between descriptive and inferential statistics?

Descriptive Statistics = Describes your population and describes findings of your subjects

Inferential Statistics = Make inferences from YOUR sample about the broader population (parametric versus nonparametric statistics)

· When do we use parametric statistics and what are the major types of parametric statistics? When and how are they used?

Your data is numeric

Your data is normally distributed

The variances are equal

You want more statistical power than non-parametric tests offer

· When do we use non-parametric statistics and what are the major types of parametric statistics? When and how are they used?

Parametric stats

-normally distributed (bell shape graph)

-quanitative data

- Use probability sampling (randomized sampling)

- Interval or ratio scale

Nonparametric stats

- non-normal distributed

-Cateforical data

-less powerful

-requires larger sample size

types of parametric statistics

t-test; paired t-test; Chi squared; ANOVA; correlation; regression

looking at the difference between 2 groups

t-test for categorical data x^2

same as t-test but with more than two groups

ANOVA when there are covariant/covariables involved

within the same person at two times

Paired t-test

T test

Chi sqaured test

ANOVA

ANOCOVA

T-test = looking at the difference between 2 groups

paired t-test = within the same person at two times

Chi squared= t-test for categorical data x^2

ANOVA= same as t-test but with more than two groups

ANCOVA = ANOVA when there are covariant/covariables involved

When to use non-parametric statistics

if we have skewed data or categorical (nominal or ordinal) data. it's less powerful and requires a larger sample size.

Types of non-parametric statistics

sign test; mann-whitney U; Kruskal-Wallis; Wilcoxon matched-pair signed rank test; chi-square goodness of fit test; odds ratio; fisher's exact test

What types of study designs are epidemiological

Disease frequency: cross sectional or surveillance

Risk factors: cohort, case control, or ecological studies

Qualitative Research Methods

Interviews – One-on-one conversations to explore experiences and opinions

Focus Groups – Group discussions to gather diverse perspectives

Observations – Watching behavior in natural settings (e.g., ethnography)

Document Analysis – Analyzing texts, media, or artifacts for themes

non-numerical data

Qualitative research Thematic, Narrative, Phenomenology, Grounded Theory, Case studies

find pattern from narrative/life stories from people

understand a broad process* about a large topic

help researchers explain what happens when people experience something common

flexible so unstructure, find themes, and make large data condensed

to understand real life stuff

Thematic analysis = flexible so unstructure, find themes, and make large data condensed

Narrative Research = find pattern from narrative/life stories from people

Phenomenology = help researchers explain what happens when people experience something common

Grounded theory = understand a broad process* about a large topic

Case studies = to understand real life stuff

· What is descriptive research and when it is used?

Research providing a detailed account of characteristics.

· What are some cautions about descriptive (observational) research? What are their strengths

· What are incidence, prevalence and mortality rate?

total number of deaths/ total number of people at risk per unit of time

reports the proportion of a population that is affected by a certain disease or condition at a given point of time (as a %).

describes disease frequency in a population (number of cases per 100,000 per ye

Incidence Rate= describes disease frequency in a population (number of cases per 100,000 per year

Prevalence Rate = The total number of cases (both new and existing) of a disease in a population at a given point or period in time so

- Prevalence= All cases (new + existing)​/Total population

Mortality Rate = total number of deaths/ total number of people at risk per unit of time (or population)

rate of disease with factor present/ Rate of disease with factor absent

The amount of risk that can be attributed to the exposure.

the actual risk of an occurrence;

The difference in absolute risk between the control and treatment groups.

an event in the exposed group to the odds in the unexposed group.

o Relative risk/odds ratios

o Absolute risk

o Absolute risk reduction

o Attributable risk

.

Absolute risk = the actual probability of an event occurring (baby basics)

Relative risk = rate of disease with factor present/ Rate of disease with factor absent

(comparing risk in 2 groups)

Odds ratio = the ratio of the group expose / unexposed (case control study so lookin backward)

Absolute Risk Reduction = The difference in absolute risk between the control and treatment groups. (how much did the treatment actually work)

Attributable risk = The amount of risk that can be attributed to the exposure.

Risk in Exposed−Risk in Unexposed then divide by total

rate of disease with factor present/rate of disease with factor absent.

If RR is greater than 1, means a positive association.

RR = 1 →

RR > 1 →

RR < 1 →

Relative risk (cohort studies)

RR = 1 → No difference

RR > 1 → Increased risk in exposed

RR < 1 → Decreased risk in exposed

an event in the exposed group to the odds in the unexposed group.

RR = 1 →

RR > 1 →

RR < 1 →

Odds ratio or relative odds

an estimate of relative risk that is calculated when the disease is rare. used with case control studie

RR = 1 → No difference

RR > 1 → Increased risk in exposed

RR < 1 → Decreased risk in exposed

The actual probability of an event occurring in a group.

the chance a specific outcome will occur so 12% of actually getting covid

Absolute risk

Difference in risk between two groups.

Absolute Risk Reduction

ARR=Risk in Control−Risk in Treatment

amount of risk that can be assigned to a particular factor

Tells how much of a disease is due to a specific exposure

How much more risk do expose people have compared to unexposed people

Risk in Exposed = Proportion of people in the exposed group who got the disease.

Risk in Unexposed = Proportion of people in the unexposed group who got the disease.

Attributable Risk

Risk in Exposed−Risk in Unexposed then divide by total

Cohort

How do you interpret a confidence intervals

1

>1

<1

1 If it includes 1 there is no association

>1 postive association so outcome is more likely in the exposed cohort

<1 outcome is less likely in the cohort

look at multiple studies and put them together, but not statisticall

combine the results of several studies and do one big statistical analysis

systematic review

Meta analysis

What does the picture represent with publication bias

publication bias

Top = no publication bias

Bottom = publication bias

open circles = non significant studies

sampling methods

every member has an equal shot of being selected

a population is divided by strata and each person in the strata has an equal chance of being selected

= an equal percentage of each strata is selected

clusters within a population are randomly selected then people are randomly selected within cluster

selecting the sample in a predetermined systemic way (every 10th person)

using only the people that responded to survey or first 20 people you see

Selection of the sample that should replicate the population and once it's met you stop sampling

pick cases that are judged to be typical of a population in which one is interested

convenience =

- Stratified random =

- Simple random =

- Quota =

- Proportional stratified random

- cluster random =

- Systematic random =

- Purposive =

Randomization

- Simple random = every member has an equal shot of being selected

- Stratified random = a population is divided by strata and each person in the strata has an equal chance of being selected

- Proportional stratified random = an equal percentage of each strata is selected

- cluster random = clusters within a population are randomly selected then people are randomly selected within cluster

- Systematic random = selecting the sample in a predetermined systemic way (every 10th person)

Not random

convenience = using only the people that responded to survey or first 20 people you see

- Quota = Selection of the sample that should replicate the population and once it's met you stop sampling

- Purposive = pick cases that are judged to be typical of a population in which one is interested

· Understand the components of each step in the survey process

1. use research question and literature

2. determine who will be sampled

3. create and test questionnaire

4. contact respondents and collect data

· What are the different types of survey questions?

categorical (yes/no)

Ordinal (drop down/ ranking)

interval (body wt, temp)

open ended

partially closed end

· What are some of the pitfalls in survey questions? Leading questions, double barreled, overlap, loaded question

what is your favorite candy

Should smart parents allow candy in the house

10-20

20-30

what do you dislike about fast food and mcdonalds

Leading Questions: These push people toward a particular answer

(Should smart parents allow candy in the house )

Loaded question: forces a response that doesn't reflect what the respondent would do (what is your favorite candy)

Overlap:

Double barreled questions: Asking two things in one question.

· What are some of the tools and methods to review a survey before sending it out?

cognitive interview

reliability testing

· What is the difference between a randomized control trial (RCT), non-randomized trial, and uncontrolled trial? What are the different types of study designs?

RCT = randomization

Non random = no random

Uncontrolled trial = no comparison (everyone gets the same experiment

observational vs experimental studies

In observational

- cohort

-cross sectional

- case control

In experimental studies

- RCT

-Non randomized

-uncontrolled trial

· What is intent to treat (ITT) and why is it important?

How is it different from per protocol

data from all the subjects who began the study should be analyzed regardless of whether they dropped out.

Important = addresses the efficacy and adherence of a treatment

per protocol = only analyze data on subjects who completed the study

· What is the importance of the number needed to treat (NNT)?

number of patients who need to be treated in order to prevent one bad or induce one beneficial outcome. Helps define the line of benefit vs. cost

The number of people who need to receive a treatment in order to prevent one additional bad outcome (like a heart attack, stroke, or death)

Heterogenity

If P=.20 ? & If I^2 is greater than 50% ? in heterogenity

measures how different the studies are from each other

If P=.20 = not significant so good there is no heterogenity

If I^2 is greater than 50% then there is heterginnity so then the P would be less than .005

If overall P is less than .005 then there is a significant difference

· What does a Kaplan-Meir curve show? Know how to read one.

Time to event or the probability of survival

attrition

When a participant drops out, or fails to complete, all parts of a study.

· What are conclusion statements and grading evidence?

??

A conclusion statement is a clear summary of what the evidence shows about a particular topic, question, or intervention.

Grading evidence is the process of evaluating the quality, consistency, and strength of scientific evidence that supports a conclusion.

· In judging existing evidence on the benefits or harms of an intervention, what is the hierarchy of study designs for evidence-based guidelines?

Systematic Reviews and Meta-Analyses

Randomized Controlled Trials (RCTs)

Cohort Studies

Case-Control Studies

Case Series

clincal experience

· Many organizations have their own grading criteria to evaluate studies in developing a guideline; however, there is a major common thread in all of them, please review each guideline to get an idea of common threads

· Why is it sometimes difficult to have strong evidence-based nutrition guidelines?

· What are the different avenues to publish data?

Technical reports,

theses/dissertations,

manuscript for publication in peer reviewed journals,

published abstracts,

posters,

oral presentations

· Review some of the examples of grammar, punctuation, units of measure and style we discussed in class.