Types of Research

Descriptive studies

“what happened”

does NOT prove cause and effect (correlation not causation)

Types:

• qualitative

• Case report, Case series, Case study

• surveys

• correlation studies or ecological

Analytic studies

“why it happened”

tests hypothesis

types:

• RCT

• experimental model

• quasi-experimental design

• cohort studies

• case-control studies

• cross-sectional studies

Measurement bias:

Hawthorne effect - effect of behaviors of individuals when being observed for research purposes

Selection bias:

Non-response bias

Attrition bias:

Loss of participants over time in a study

Publication bias:

When researchers see publication studies only if they contain statistically significant results

Instrument bias

Scales used to give higher or lower readings

Recall bias

Participants cannot recall

Anything to do with self-reporting

Social Desirability bias:

Participants choose answer that will be viewed favorably

Extreme response set bias:

When respondents pick the more extreme answers

Acquiescence bias

When respondents are more likely to agree with a list of statements than disagree

Alternative hypothesis (H₁)

• States that there is a real effect, a significant difference, or a true relationship between variables in a population. This is the claim the researcher wants to test or prove.

Ex.

Null hypothesis (H₀)

• States if there is a difference in the dependent variable, it is due to chance (luck) and is not due to the independent variable

• We hope to reject because then it would NOT be due to chance

Ex.

Type I Error (alpha):

When you reject the null hypothesis when it is true

Ex.

Type II Error (beta):

When you accept the null hypothesis when it is false

Ex.

IRB - Institutional Review Board

• under FDA

• committee established to review and approve research involving human subjects to ensure it is conducted within all ethical and federal guidelines.

• It may also be known as (IEC) Independent Ethics Committee, (ERB) Ethical Review Board, or (REB) Research Review Board.

Steps in research

1. Identify a relevant and important topic: review published research literature related to the topic

2. Develop well-considered research question (who, what, how)

   1. clear, simple statement in a few words, in a complete grammatical statement

3. Research question leads to a hypothesis

   1. measurable, specifies population being studied, time frame, type of relationship being examined, defines variables being studied, states level of statistical significance

   2. hypothesis is a prediction of a relationship

   3. null hypothesis

   4. hypothesis should be feasible, interesting, novel or innovative, ethical and relevant

      1. Consider PICO (Population, Intervention, Comparison, Outcome)

   5. Prepare research protocol; methodology to solve the problem

   6. Organize methods and materials, collect and analyze data

   7. Study results and make decisions

   8. Study design and checklists

relevance or validity

ability to measure phenomenon it intends to measure

internal validity

tests whether the difference between the two groups is real

(has the experimental group really performed differently?)

external validity

tests whether the or not a generalization can be made from the study to a larger population

ANOVA - analysis of variance

• tool used to evaluate validity

  • asks whether the difference between samples is a reliable one that would be repeated

  • used when several (three or more) products compete against one another

  • compares the variance between groups with the variance within groups

  • answers: are there one or more significant differences anywhere among the samples

reliability

• consistency or reproducibility of test results

• test, the retest later; are results similar?

• parallel forms: two separate but similar forms of the same test at the same time. Reliability determined by the degree to which the sets of scores coincide

• split halves: divide the test in half. Reliability is determined by the degree of similarity of results

• precision - amount of variation results in greater precision in the measurement and greater reliability

sensitivity and specificity

use if protocol involves screening for a particular condition; evaluates the cut-off value being used

sensitivity

proportion of afflicted individuals who test positive

specificity

proportion of non-afflicted identified as non-afflicted

variables

characteristics that may have different values from observation to observation

nominal (non-ordered) variable

variables that fit into a category with no special order

• ex. gender, race, marital status, present or absent

rank order (ordinal scale) variable

observations compared with each other ndput in order, perhaps from best to worst, or state of disease from 1 to 4

numerical discrete variable

data with numbers

• ex. number of clinic visits

numerical continuous variable

underlying continuous scales

• ex. blood pressure

Probability

each segment of the population will be represented in the sample

• selects units from a much larger population

• uses randomization - select a sample from the whole population so the characteristics of each of the units approximates the characteristics of the whole population

Non-probability

no way of forecasting that each element in the population will be represented in the sample

• convenience or accidental - take units as they arrive on the scene - no attempt to control bias

• quota - select units in the same ratio as they are found in the general population

Correlation (r)

• linear correlation coefficient (r)

• the closer the points are to the line, the stronger the degree of linear relationship

• value of “r” is always between -1 and 1

• r=1 when all points lie exactly on a straight line with a positive slope

• r=-1 when all points lie exactly on a straight line with a negative slope

• r=0 there is no linear relationship, no agreement

p value - clinical significance

p ≤ .05 significant difference, results are reliable

p < .01 very significant difference, more reliable results

p < .0001 very, very significant, reliable results

p > .05 not very significant difference; not reliable results

descriptive statistics

summarizes and describes aspects of a set of data

inferential statistics

techniques that allow conclusions to extend beyond an immediate data set; what is the probability that the results can be applied to a larger group; what can you infer from the results of your study