Using Computers for Data Analysis

1
Using Computers for Data Analysis

• Adam Schlichting
• University of Illinois at Chicago
• Department of Emergency Medicine

2
IntroductionWhy?

• Too many calculations to do on a handheld
  calculator

3
IntroductionPrograms

• EpiInfo
• Centers for Disease Control and Prevention (CDC)
• Free software
• http//www.cdc.gov/epiinfo/

4
(No Transcript)
5
IntroductionPrograms

• Statistical Program for Social Scientists (SPSS)
• Easy to use, point and click
• Similar to Microsoft Excel
• Fairly powerful

6
(No Transcript)
7
IntroductionPrograms

• Statistical Analysis Software (SAS)
• Very powerful
• Not so easy to use

8
(No Transcript)
9
IntroductionPrograms

• Other Programs
• PEPI
• STATA
• FOCUS

10
IntroductionPrograms

• Well focus on
• SPSS
• EpiInfo for special situations
• Easiest to use
• Tell you everything you need to know 99 of the
  time
• Biostatisticians exist for the
  remaining 1

11
SPSSThe Program
12
SPSSThe Program
13
SPSSImporting Data

• Excel is easier to enter and manipulate data
• Need to import data
• Excel
• Access
• DBase
• Delineated text

14
SPSSImporting Data Specify File Type
15
SPSSImporting Data Specify File Location
16
SPSSImporting Data Import Variable Names
17
SPSSImported Data Complete
18
SPSSSaving Imported Data as a SPSS File
19
SPSSSaving Imported Data as a SPSS File
20
SPSSAnalysis Frequency Counts

• Do frequency counts of everything
• Points out errors that need to be cleaned
• Look for obvious mistakes
• Age 650 instead of 65

21
SPSSAnalysis Frequency Counts
22
SPSSAnalysis Frequency Counts
23
SPSSAnalysis Frequency Counts

• Use shift/click and Ctrl/click to select
  variables

24
SPSSAnalysis Frequency Counts
25
SPSSAnalysis Frequency Counts
26
SPSSAnalysis Frequency Counts
27
SPSSAnalysis Frequency Counts Printing
28
SPSSAnalysis Frequency Counts Printing
29
SPSSAnalysis Central Tendency

• Useful for demographics, lab values
• Defaults
• N
• Range
• Mean
• Standard Deviation

30
SPSSAnalysis Central Tendency
31
SPSSAnalysis Central Tendency

• Use shift/click and Ctrl/click to select variables

32
SPSSAnalysis Central Tendency
33
SPSSAnalysis Central Tendency
34
SPSSAnalysis Central Tendency
35
SPSSAnalysis Crosstabs

• Compare subgroups of catagorical variables on
  other variables
• Not used for continuous variables
• Our example
• In this sample, does outcome differ by
• prehospital GCS?
• sex?

36
SPSSAnalysis Crosstabs
37
SPSSAnalysis Crosstabs
38
SPSSAnalysis Crosstabs
39
SPSSAnalysis Crossabs
40
SPSSAnalysis Crossabs
41
SPSSAnalysis Crosstabs

• Gives a nice breakdown of data
• But how do we know if a relationship exists?

42
SPSSAnalysis Crosstabs

• Use Statcalc in EpiInfo to quickly calculate Odds
  Ratios, Relative Risks, Confidence Intervals and
  p-values

43
EpiInfoStatcalc
44
EpiInfoStatcalc
45
EpiInfoStatcalc
46
EpiInfoStatcalc

• Must translate orientation in SPSS table into
  Exposure/disease orientation for Statcalc
• Disease outcome
• Exposure risk

47
EpiInfoStatcalc

• Disease outcome Alive
• 1 alive, 28 days
• 2 dead, 28 days
• Bad outcome (death) disease
• Exposure risk sex
• 1 male
• 2 female

48
EpiInfoStatcalc
49
EpiInfoStatcalc
Nothing significant
50
EpiInfoStatcalc

• If we had other data

51
SPSSBack to SPSS
52
SPSSAnalysis Regression

• When several variables work together, which plays
  the most important role in predicting an outcome?
• How well does this model predict outcome?
• Short Example
• In predicting outcome in a trauma study, which of
  the following factors is most important in
  predicting outcome gender, pre-hospital GCS, age?

53
SPSSAnalysis Linear Regression
54
SPSSAnalysis Linear Regression

55
SPSSAnalysis Linear Regression
56
SPSSAnalysis Linear Regression
57
SPSSAnalysis Linear Regression
58
SPSSAnalysis Logistic Regression
59
SPSSAnalysis Logistic Regression
60
SPSSAnalysis Logistic Regression
61
SPSSAnalysis Logistic Regression
62
Computer Statistical AnalysisSummary

• Brief overview of
• Importing data
• Frequency counts
• Central tendency
• Crosstabs and ORs/RRs (EpiInfo)
• Linear regression
• Logistic regression

63
Computer Statistical AnalysisSummary

• Dont be afraid to ask people to help

64
Questions?