Statistical Fridays

1
Statistical Fridays

• J C Horrow, MD, MSSTAT
• Clinical Professor, Anesthesiology
• Drexel University College of Medicine

2
Session Review

• Sensitivity / specificity
• Predictive value
• Effect of disease prevalence
• The ROC curve

3
Session Overview

• Learn new concepts
• Observations vary
• Observational vs. experimental data
• Graphing your data
• Example
• 50 patients induced with TPL or PPF

4
Observations Vary
Subject JHs systolic BP is 151 mmHg. He applies
1 mL of 5 minoxidil to his scalp in a vain
attempt to keep his remaining hair. 15 min
later, his systolic BP is 148 mm Hg.
Q Did his BP decrease ?
Q May we state BP decreased ?
5
Observations Vary
6
Observations Vary

• When do things differ?

7
When do things differ?

• When statistical tests indicate so.
• Data nearly always differ numerically
• Only statistics tells us when numbers differ.
• There is no such thing as a difference that is
  not a statistical difference.

8
When do things differ?

• Only when they differ statistically.

9
Population v. Sample

• POPULATION
• The theoretical cohort about which we wish to
  draw conclusions..
• Examples
• Patients with heart disease
• Pregnant women with hypertension
• Patients with antithrombin deficiency

1
10
Population v. Sample

• SAMPLE
• The specific subjects for whom we have
  measurements
• Patients with heart disease
• 25 patients with classic angina
• Pregnant women with hypertension
• 45,420 pregnant women taking Atacand
• Patients with antithrombin deficiency
• The patient seen in clinic yesterday

11
Population v. Sample Models

• Statistical Model
• Device by which we infer properties about a
  population based on information obtained in a
  sample
• ALL MODELS ARE WRONG.Some are useful.

12
Statistical Model

• Example
• Let x TPL dose (mg/kg) at induction
• Let y decrease in SPB at induction

y b0 b1x e
13
Types of Data

• Observational
• Experimental

14
Types of Data

• Observational
• Experimental

ASSOCIATION
Fifty (50) patients receive TPL or Propofol for
induction. The ?SPB for each is recorded.
CAUSATION
Fifty (50) patients are randomly assigned to
receive either TPL or Propofol for induction. The
?SBP for each is recorded.
15
Graphing Data

• Histograms
• Scatterplots
• Boxplots

16
Example SBP data

• 100 measurements total
• 25 SBP before TPL
• 25 SBP after TPL
• 25 SBP before PPF
• 25 SBP after PPF
• Calculate DSBP for each

17
HISTOGRAM
18
Histogram by Treatment
19
Scatterplot before v. after
20
Scatterplot by treatment group
21
Boxplot of DSBP by treatment
22
(No Transcript)
23
(No Transcript)
24
Session Review

• New concepts
• Observations vary
• Observational vs. experimental data
• Graphing your data
• Example
• 50 patients induced with TPL or PPF
• Homework
• 20 patients given spinals for C-section