JoinPoint

1
JoinPoint

• How to trend VistaPHw output
• September 20, 2002

2
Goals for today . . .

• What is JoinPoint?
• Why use JoinPoint?
• How does JoinPoint work?
• How does JoinPoint work?

3
What is JoinPoint?

• Developed by SRAB/NCI
• Statistical method tool
• Used for all NCI trends
• Identifies trends and changes in trends
• Free

4
Why Use JoinPoint?

• Because it identifies trends and changes in the
  trends.
• Because the current test for trend in VistaPHw
  makes assumptions about the data that may or may
  not be true and generates results which may or
  may not be true.

5
How does JoinPoint work?
6
Back in the old days . . .
7
. . . but is that the pattern we see?
8
JoinPoint - 10 year trend
Is there any single trend line or combination of
trend lines that fit the data better than this?
9
JoinPoint best fit (1 joinpoint)
10
JoinPoint best fit (1 joinpoint)
2.3 per year increase 1992-1998
no significant trend 1998-2001
11
Maternal smoking
12
Infant death rates
13
Low birth weight rates
14
Late or no prenatal care
15
Late or no prenatal care
16
How does JoinPoint work?

• Fire up VistaPHw
• Generate JoinPoint output from VistaPHw
• Fire up JoinPoint
• Import the VistaPHw output into JoinPoint
• Run trend analysis in JoinPoint
• Review JoinPoint output
• Wash, rinse repeat with modifications

17
Well be using . . .

• Developed for the Health of Washington State
  Report
• Provides guidance on JoinPoint settings
• Talks to the Bonferroni adjustment and the
  maximum number of Joinpoints

18
Exercise 1 Using the simple output

• In VistaPHw choose
• All cause deaths
• Age-adjusted (2000)
• Total Race
• Total Sex
• 1980-2003
• State Total

• In Launch Panel
• Sort by Year
• Check JoinPoint Code
• Check Simple
• Check Text Output
• Save on Desktop

19
Exercise 1 Using the simple output

• Start-up JoinPoint
• Under File click New Session
• Check Other Data Files
• Find your VistaPHw output file press open
• Confirm in Data Preview press Okay

20
(No Transcript)
21
Exercise 1 Using the simple output

• For Dependent Variable press the drop-down button
  and choose RATE
• For Independent Variable press the drop-down
  button and choose YEAR
• For Standard Error press the drop-down button and
  choose SE
• Press the lightning bolt

22
(No Transcript)
23
Exercise 1 Using the simple output

• To export from JoinPoint
• Press Output (upper left hand corner)
• Choose Export as Text . . . Data
• Use Browse to choose location
• Name your output
• Press Okay

24
Exercise 1 Using the simple output

• To export from JoinPoint
• Press Output (upper left hand corner)
• Choose Export as Text . . . Data
• Use Browse to choose location
• Name your output
• Press Okay

25
Break

• Be back in 10 minutes . . .

26
Exercise 2 The complex output

• In VistaPHw keep all the choices the same as
  before, except
• Select all races and Hispanic ethnicity
• Select males and females (and total)
• Press Launch
• Sort by
• Race (or Sex)
• Sex (or Race)
• Year
• JoinPoint Code All will be checked Press Run
  Now

27
Exercise 2 The complex output

• In VistaPHw keep all the choices the same as
  before, except
• Select all races and Hispanic ethnicity
• Select males and females (and total)
• Press Launch
• Sort by
• Race (or Sex)
• Sex (or Race)
• Year
• JoinPoint Code All will be checked Press Run
  Now

If you sort by YEAR first, the first records in
your output would be all 1980 1980, total sex,
total race 1980, male, total race 1980, female,
total race 1980, total sex, white . . .
28
Exercise 2 The complex output

• In JoinPoint
• Under File click New Session
• Check Other Data Files
• Find your VistaPHw output file press open
• Confirm in Data Preview press Okay

29
Exercise 2 The complex output

• For Dependent Variable press the drop-down button
  and choose Z_RATE
• For Independent Variable press the drop-down
  button and choose Z_YEAR
• For Standard Error press the drop-down button and
  choose Z_SE
• DO NOT press the lightning bolt, yet

30
Exercise 2 The complex output

• Defining the By Variables in JoinPoint
• Below the Variable Name box, press Add...
• Choose Z_SEX and press Okay
• Click on Z_SEX in the Variable Name box
• Press Define
• In the box under ValueLabel type
• 0total
• 1male
• 2female
• Press Okay

31
Exercise 2 The complex output

• Below the Variable Name box, press Add...
• Choose Z_Race and press Okay
• Click on Z_Race in the Variable Name box
• Press Define
• In the box under ValueLabel type
• 0total
• 1white
• 2black
• 3am ind
• 4asian
• 6hisp
• Press Okay

32
Exercise 2 The complex output

• Below the Variable Name box, press Add...
• Choose Z_Race and press Okay
• Click on Z_Race in the Variable Name box
• Press Define
• In the box under ValueLabel type
• 0total
• 1white
• 2black
• 3am ind
• 4asian
• 6hisp
• Press Okay

Now hit the lighting bolt
33
(No Transcript)
34
(No Transcript)
35
Break marvel at the trends

• Reconvene in 5 minutes?

36
Whats going on with the Hispanic trends???
37
Exercise 3 On your own . . .

• Run trends by sex and race for all deaths from
  1990 to 2003
• After completing 1 (above) run again resetting
  the maximum number of joinpoints from 3 to 2

38
Exercise 3 On your own . . .

• Run trends by sex and race for all deaths from
  1990 to 2003
• After completing 1 (above) run again resetting
  the maximum number of joinpoints from 3 to 2

39
American Indian males, 1990-2003

• Maximum joinpoints3

• Maximum joinpoints2

Why are they different??
40
American Indian males, 1990-2003

• Maximum joinpoints3

• Maximum joinpoints2

Bonferroni adjustment
41
Bonferroni adjustment

• The Bonferroni adjustment is used to control the
  error probability of each of the multiple tests
• a1a/(MAX - MIN)
• a0.05
• If the individual test p-value is less than a1,
  the null is rejected
• In other words, the test for significance is more
  stringent with 3 joinpoints than with 2

42
How many joinpoints should I set as the maximum??

• In general, there must be at least
• 5 years of data to specify 1 joinpoint,
• 8 years for 2 joinpoints,
• 11 years for 3, and
• 14 years for 4
• We strongly recommend selecting one fewer than
  the maximum number of possible joinpoints

43
Feel free to play with the data

• Ill wander around and answer any questions.