Reading Research Articles how to approach statistical information

1
Reading Research Articleshow to approach
statistical information

• University Graduate School
• Marion Haas
• March 18, 2008

2
Introduction

• Please interrupt me!
• In this session
• Aim
• Approach to adopt
• Practical skills to learn
• Type of statistical information presented
• How to read articles
• Practice!

3
Aim of this session

• Recognise diversity among students
• Often not mathematical
• Different notions of statistics
• Translate concepts into understandable language
• Approach levels of statistics in a step-wise way
• Knowing where to look in the text
• Interpreting the text
• Explaining the text in your words

4
Definition

• Statistics is a mathematical science pertaining
  to the collection, analysis, interpretation or
  explanation, and presentation of data. It is
  applicable to a wide variety of academic
  disciplines, from the natural and social sciences
  to the humanities. Statistics is also used for
  making informed decisions in government and
  business.

5
Statistics

• Often first contact with numbers beyond simple
  concepts at school
• Different way of using numbers
• All future researchers professionals in many
  areas need
• Working knowledge of how to understand
  statistical output
• Some sophisticated understanding

6
Approach (1)

• Develop and broad integrated view of statistics
• Use the text know where to look
• Interpret the text what does it mean?
• Explain the text use your words
• Ask a series of questions
• Develop critical reading skills
• Understand the statistical content of articles
• Connecting topic to professional lives

7
Approach (2)

• 3 types articles
• Elementary descriptive statistics techniques
• Introductory techniques of statistical inference
• Advanced statistical techniques
• 2 type of reading
• Preliminary the same for all articles
• In-depth higher order skills used to
• Generate text analyses
• Place text in context

8
Preliminary reading

• General skim to decide if useful
• Title, author, abstract, first last paragraphs,
  diagrams, graphs, tables
• 5 minutes, 5 dot points
• If continuing
• Questions about aim, audience, context
• Understand main points of article

9
Specific preliminary questions (1)

• What is the research question?
• How is it presented?
• Why is it presented in this way?
• What research methods are used?
• Observational, experimental, neither
• Sampling techniques?

10
Specific preliminary questions (2)

• What data are used?
• How are they dealt with?
• How are they presented/described?
• Statistical techniques?
• List
• Why is each one used?

11
Applied statistics

• Descriptive statistics
• Statistical methods used to summarize or describe
  a collection of data
• Inferential statistics.
• Modeling patterns in the data to account for
  randomness and uncertainty in the observations,
  used to draw inferences about the process or
  population being studied
• Both descriptive and inferential statistics
  comprise applied statistics

12
Practice

• Lets look at the first article
• Veal, Tony (1997). Gambling trends in the 1990s.
• Skim for 5 minutes
• First and last paragraphs
• Diagrams, graphs, tables
• What sort of statistical work (activity) has
  Veal undertaken?
• What type of language has he used?

13
Descriptive statistics

• Used to summarise data
• Numerically or graphically
• Describe the sample
• Numerical
• Mean, median, mode, range, standard deviation etc
• Graphical
• Charts and graphs

14
Veal in depth

• Aim and audience
• What is the main aim
• What clues do the statistical techniques give
  about the aim?
• What audience is it written for?
• Content
• Figure 1 In 1996, which was the largest
  component of gambling expenditure?
• Which was growing fastest?
• Why is NSW the most mature gambling market?

15
Veal (2)

• Analysis
• Do the graphics work here? Why?
• How could they be improved?
• Is the information presented in an objective
  fashion? Does the author approve of gambling?
• Is the writing style formal or informal? Examples?

16
Inferential statistics

• Used to
• Model patterns in data
• Account for randomness
• Draw inferences about the larger population
• Inferences
• Hypothesis testing (answer yes/no)
• Make estimates of numerical characteristics
  (estimation)
• Describe associations (correlation)
• Model relationships (regression, ANOVA, time
  series)

17
Important concepts (1)

• Correlation
• When two characteristics tend to vary together,
  as if connected
• Eg income and age at death
• Poorer people die younger
• BUT, does this mean that poverty causes death or
  that poor health causes poverty?
• Correlation DOES NOT imply causation

18
Important concepts (2)

• Probability
• Fundamental concept used to understand randomness
• Methods able to estimate and correct for
  randomness in design, sample, data collection
• Alpha level refers to probability of Type 1 error
• Significant association found when does not exist
• Alpha level conventionally set at 0.05 ie if
  plt0.05 we accept that a significant association
  exists
• 95 confidence intervals (CI) give more
  information 95 CIs are significantly different
  if DO NOT overlap

19
Statistical methods

• Usually asking
• What will happen to response (outcome, dependent)
  variables if changes occur in predictor
  (independent) variables?
• Experimental observational studies
• Both investigate how changing independent
  variables affects behaviour of dependent
  variable(s)
• Difference lies in HOW the study is conducted

20
Experimental study

• Basic design
• Initial measurement
• Change conditions
• Measure again
• Control group
• Blind measurement

21
Observational study

• Basic design
• Data gathered
• Correlations between predictors (independent) and
  response (dependent) variables investigated
• Control

22
Practice

• General questions
• Viney et al Skim for 5 minutes
• What is the research question
• What research methods are used?
• What data are used?
• What statistical techniques are used?

23
In-depth questions (1)

• Aim and audience
• What is the main aim?
• What clues do the statistical techniques give
  about the aim?
• What audience is this aimed at? Reasons?
• Content
• What is a randomised trial?
• What does eligibility criteria mean?

24
In-depth questions (2)

• Analysis
• Who were the subjects for this trial? how were
  they selected?
• there was no significant difference between the
  arms in the number of thoracotomies avoided
  (Table 3, p0.2). Can you explain this in lay
  terms?
• Can you think of some reasons that the management
  of NSCLC did not change as much as it could have?

25
More practice

• Vignaendra and Fitzgerald
• What is the research question
• What research methods are used?
• What data are used?
• What statistical techniques are used?

26
In-depth questions (1)

• Aim and audience
• What is the main aim?
• What clues do the statistical techniques give
  about the aim?
• What audience is this aimed at? Reasons?

27
In-depth questions (2)

• Content
• What do the authors mean by caution cohort and
  conference cohort?
• Explain what is meant by bi-variate and
  multi-variate analysis
• Table 2 what characteristics (variables) are NOT
  significant? Why is this important?

28
In-depth questions (3)

• Analysis
• Look at the logistic regression model/s (Tables 2
  and 6)
• What is the response (dependent) variable?
• What are the predictor (independent) variables?
• Can you explain these results in lay terms
• What does survival analysis mean in this context?
• What of young people survived at least 12
  months before being seen in court again? (Figures
  3 4)
• By what time following the first conference had
  50 of young people been seen in court again?
  (Figure 4)

29
Reference

• Leigh Wood and Peter Petocz (2003). Reading
  Statistics, Mathematics Study Centre, University
  of Technology, Sydney. Printed by UTS Printing
  Services