Statistics: What do I need to know?

1
Statistics What do I need to know?

• What are Chi squares, t-tests, ANOVA,
  correlations and regressions?
• How do I know if the researchers used the correct
  statistical tests?

2
How do statstics relate to an indivudal
participant

• Difference between climate and weather
• Can predict exactly what this person will do!

3
Participants 65 Pregnant Adolescents in Teen
Parent Programs

• Means age 16.12, GPA 2.04, age/grade lag .17,
  Grade 10.8, FOB/MOB age difference 3.17 year
• Percentage AA (13), Hispanic (42), Caucasian
  (45).
• Refusal rate 2
• Attrition 6 (fetal demise, diagnosis of CA)
• 7 Sites Relationship of demographics to school
  attendance site, ethnicity, SES, age, grade,
  GPA
• P values .13 - .95
• Was I happy?

4
Statistical tests Were the right ones used?

• Demographics 7 groups ANOVA used.. not 42
  T-Tests
• Linear Regression Not single order correlations.
  Symbol is r
• DV school attendance interval data 1-21 days.
• Variables entered into the computer in a
  step-wise manner based on the TPB
• 1st Demographics then TPB concepts

5
Theory of Planned Behavior
Demographics TBP Constructs Intention predicts behavior Outcome
Age SN
School success PC Intention Behavior
SES GPA Fx Hx Attitude
6
What are correlations?

• The relationship between the IV (x axis) and the
  DV (the y axis)
• R .6 for every 1 on the Y axis the x axis line
  goes up .6

7
What does Regression mean?

• All the little dots are each data point (i.e each
  score)
• r refers to how much y (DV) changes in
  relationship to X (IV)
• The solid line is the line of best fit

8
Ven Diagrams and correlations

• IV Red (attitude)
• IV Blue (social norm)
• DV Yellow (attendance)
• Orange what attitude contributes uniquely to
  attendance (beta wt)
• R2 orange white green

9
DV C School attendance IV A Attitude r
AC ABC R2 ACbeta AC IV B Social
Norm r BCABC beta BC

• Visualizing Shared and Unique variance How much
  do we understand about the DV?

10
Single order correlations can be deceiving!
IVs beta F p
Attitude .006 .96
Perceived Control .098 .45
Social Norm .159 .17
Intention .234 .63
Full Model 3.07 .023
A PC SN I
School (DV) .06 (.67) .25 (.09) .29 (.04) .33 (.01)
A .10 (.52) .00 (.99) .06 (.69)
PC .28 (.05) .38 (.00)
SN .22 (.10)
11
Did the theory (model) work? How do you
know?Why are the R2 and Adjusted R2 different
(think sample size!)
R2 Adjusted R2 F P (.05)
.140 .093 3.007 .023
12
Does the TBP help us understand school attendance?

• IVs Demographics Did not predict school
  attendance
• P .28 - .62
• IVs Attitude Social Norm Perceived Control
• Intention
• Predicted
• DV School Attendance
• Why is this a helpful thing to know?

13
Confidence Intervals

• If crosses 0 or 1 then results are not
  significant
• The larger dot is the mean
• The line relates SD p value (p .05 then line
  .95)

14
Effect Size Chi Square

• Small Effect size Dont smoke CA
• Med Effect size smoke CA
• Lge ES smoke/emphsyema/fx hx

Df N-1 Sm Med LGE
1 795 87 26
2 964 107 39
3 1090 121 44
15
Effect Size t test/ANOVA

• Small Effect
• Med Effect
• Lge Effect

Df N-1 Small Med Large
2 393 67 26
3 322 52 21
4 274 45 18
16
ANOVA More than 2 groups

• Where is the difference??
• Post Hoc will tell you!
• i.e. Significant difference between groups 1 3
  and 1 4 3 4

Group X Minutes of exercise X Weight loss (lbs) P value
1 0 -.1 .04
2 30 .2 .02
3 15 .3 .05
4 45 .8 .01
17
Did they use the right test?

• Why we need p values
• Chi Square comparing two or more groups using

18
Did they use the right tests means

• T test comparing the means of two groups
• ANOVA comparing the means of three or more groups
• Did they do a post hoc test

19
Did they do the right tests comparisons

• Correlations comparing two variables ( 1 IV 1
  DV) on a continuum
• Regression there is more then one IV and there
  is one DV
• IV 1 IV2 IV3 DV

20
P values and fishing expeditions

• What does a p value of .05 mean?
• So if I do 100 comparisons.. How many will be
  related by chance alone?

21
How much confidence should I have in statistics?

• Statistics dont lie.. Liars use statistics
• Based on what I know about this subject (people,
  disease) does this make sense
• You can have statistical significance, but not
  clinical significance, but not the other way
  around!