Biostatistics and Research

What is Research

“to go about seeking”

What does the process do?

Collects and analyzes information/data

The goal:

increasing our understanding of the world

Research Process steps

1.Observe 2.Identify a research problem 3.Review the literature 4.Develop a hypothesis 5.Collect data 6.Analyze/Interpret data and draw conclusions 7.Report the research

Research topic:

board

Research problem:

general issue, concern, etc

Purpose:

major intent or objective of the study

Research question:

the specific thing being addressed

Literature Review

Summary of published work in a field. Scholarly books, journal articles, etc. Helps you understand where the gap is

Hypothesis

A statement about how the world works. Can be tested and disproven

Statistics:

“technology that describes and measures aspects of nature from samples”

The goal:

estimation

“Assess differences between

groups and relationships between variables”

The goal: estimation

 “Assess differences between groups and relationships between variables”

Parameters:

quantity describing a population

Understanding statisitics

Prepare you to do research and to evaluate studies and statements

What is a scientific paper?

A way of reporting findings to a specific audience. The output of your study. Reflects the research methods

Title

Concise: Max 15-ish words

 Clear, understandable by audience

 No question marks and exclamation points

 Keywords! Yes!

 No abbreviations, jargon, etc

 Follow journal guidelines

 Make your findings clear

 Title must be substantiated in the article

Abstract

Short summary of:

 Aims

 Main findings

 Quick overview of what was done and what was interesting about it

 “shop window” for your paper

Introduction

The literature review! “Written summary… that describes the past and current state of information” (Creswell 2008)

Materials and Methods

What you did. Often organized into subheadings. Appropriate level of detail

Results

What you found. Often organized into subheadings. Includes statistics. Does not discuss the implications of your results

Discussion

Back to your hypothesis! Interpretation and reasonable speculation. No new data or analysis

Referecnes

In-text citations, reference list at the end. Follow the guidelines of the journal you submit to

Watch your Abbreviations

Purpose of the literature review

Convince your reader that you understand the state of the literature on your topic. Lead the reader to your hypothesis. Provide evidence that your study is necessary. Build your research skills as an investigator who can follow the leads in the literature. Learn how other researchers carried out their work

Primary source

“Immediate, firsthand account”. Original research. Adds directly to our knowledge of a topic

Secondary source

One step removed. Quote or use primary sources. Add interpretation and analysis. Compile what we currently know about a topic. A review article

Primary:

Experimental methods. “we demonstrate/find/etc”. Figures that show data directly

Secondary

Will often be called a “review”. Does not claim to have demonstrated anything. May reference how they searched for sources

Writing a good research question

Testable. Begins with how, what, when, why, where. Should be comparing things, looking for relationships

Hypothesis

Supposition about how the world works. Potential explanation of observed phenomena. Follows from observations and research questions. Tentative and unproven. TESTABLE

Theory

Explanation of observed phenomena. Broad. Follows from many studies/experiments/evidence. Has a great deal of support from the data

Prediction

Something that ought to happen if your hypothesis is correct

Null hypothesis (H0):

“There is no relationship between these variables” or “These two groups are from the same population”

Alternative hypothesis (HA):

“There is a relationship between these variables” or “These two groups are from different populations”

Variable:

"any characteristic or measurement that differs from individual to individual“

Data:

“raw measurements of one or more variables from a sample of individuals”

Types of Variables

Categorical and Numerical

Categorical Variables

Nominal and Ordinal

A nominal variable

sorts data into distinct, named groups or categories that have no inherent order, rank, or numerical value. Example color

An ordinal variable

values that have a natural, logical order or ranking, but no measurable or equal distance between the categories. Example: size class

Numerical

Discrete and Continuous

Discrete

Values are obtained by counting. They have clear, distinct spaces or gaps between values (usually whole numbers)

Continous

Values are obtained by measuring. They can take any value along a continuum, including decimals and fractions

Independent Variable

explains or predicts the dependent variable

Dependent Variable

the variable that can be predicted. What you are trying to study

Confounding variables

Variables you were not trying to encounter

Observational studies:

IVs explain, predict, or correlate with DV

Randomized experiments:

Researcher controls IV, measures DV. IV causes changes in DV. Treatment group = always IV

Precision:

“spread of estimates resulting from sampling error”

Accuracy:

“average of estimates is centered around the true population value

Types of studies

Observational and Experimental studies.

Observational study

Observing a given sample and Not trying to affect them or their environment

Observational study Advantages:

Allows us to observe the natural environment and Does not require randomization

Observational study Disadvantages

Cannot prove cause and effect and Often affected by other variables

Sampling unit:

Objects/individuals/etc from which measurements are taken

Sampling unit: Characteristics

Need to be independent/avoid pseudoreplication. Results in one data point. Should be consistent throughout the studyy

Sampling Unit types

Haphazard, Systematic, Stratified Random and Random

Haphazard

Arbitrarily select samples, Highly subject to bias, AVOID

Random

Ideal, Unbiased, Uses RNG, Time Consuming and Inconvenient

Systematic

Samples at regular intervals and Possible bias

Stratified random

ID possible confounding variables. Select random samples from each group. Hybrid between systematic and random

Experimental Study

Compare how individuals of the same type behave in different environments and Compare how different types of individuals behave in a given environment

Experimental Study advantages

Proves cause and effect and You have control

Experimental Study disadvantages

Need to control confounding variables. Need random sampling. Artificial environment

Experimental units

Objects/individuals/etc to which IV is applied

Experimental units characteristics

DV measurements are taken. Should be independent of each other. Results in one data point. Should be consistent throughout study

Randomized Block Design characteristics

Useful if environment/population varies in a systematic way. Divide samples into blocks (e.g. elevation, gender, disease). Assign treatments randomly within blocks

Factorial Design

2 or more treatments with different levels. Must look at all combinations

Pseudoreplication

Violates the assumption of independence. Makes a false claim about how many independent samples we have. P-values that are too small

WHAT IS A GRAPH?

A way to display data

Tables:

IV and DV both categorical, Very few data points. The audience needs to see every data point

Graphs:

2 IVs. Statistics. Many data points

Bar graph rules:

Start at zero. Equal width bars. Do not fuse bars. Give sample size in figure legend

Mean, Median, Mode

Mean: arithmetic mean/average. Median: “middle” value. Mode: most common value

COMMON DESCRIPTIVE STATISTICS FOR BAR GRAPHS/BOX PLOTS

Mean, Standard Deviation, Standard Error, Median, and Interquartile Range

Mosaic plot:

uses relative frequency as height and width is proportional to number of observations

CATEGORICAL DATA

Bar graph vs. box plot

GROUPED BAR GRAPHS AND MOSAIC PLOTS

2 or more categorical variables

NUMERICAL DATA

Frequency table and Histogram

Histogram:

uses rectangular bars to display frequency. Data values split into bins. Frequency of values falling into each bin is graphed

Skew:

asymmetry in the shape of a frequency distribution for a numerical variable

Right skew:

tail extending to the right

Left skew:

tail extending to the left

A bimodal distribution is a

probability distribution or data set that has two distinct peaks, or modes, where data points cluster most frequently

Outliers

Common. May result from mistakes or real features of nature. Do not discard unless you can prove it’s an error

RULES FOR HISTOGRAMS

Start at zero. No separation between bars. Use readable numbers for breakpoints. Give n in figure legend

How many intervals to use?

• Sturges’s Rule of Thumb? 1+ln(n)/ln(2). N=number of observations

RELATIONSHIP BETWEEN 2 NUMERICAL VARIABLES

Scatter Plot. Line Graph. Map

SCATTER PLOT

Each observation = 1 point on a graph with 2 axes

X axis position:

measurement of explanatory variable

Y axis position:

measurement of response variable

Results in a cloud of points that can

reveal relationships

LINE GRAPH

Displays trends in time or some other ordered series. Points connected by lines

MAP

Spatial equivalent to a line graph

Map Explanatory variable:

points in space

Map Response variable:

indicated by color

PRINCIPLES OF EFFECTIVE DISPLAY

Show the data. Represent magnitudes accurately. Draw graphical elements clearly. Make displays easy to interpret