Vous êtes sur la page 1sur 25

A radical view

on plots
in analysis
Hein Stigum
Presentation, data and programs at:
http://folk.uio.no/heins/

Agenda

09-11

Why use graphs


Common use
A graphic evolution
Plots in analysis

H.S.

Why use graphs?

Problem example
Lunch meals per week

30
0

10

20

Percent

40

50

Table of means (around 5 per week)


Linear regression

09-11

3
4
5
Lunch meals per week

H.S.

Problem example 2
Iron level
Linear regression: Males 9.4 units higher iron level
Logistic regression: Males 10.4 times more anemia

50

09-11

cutoff

90

Male mean

Female mean

Iron level by sex

115125
Iron level

H.S.

150

190

Problem example 3
Weight on blood pressure
Variable

15.632724
10.944734

Adjusted
20.618903
9.67202
-28.388124

Model

Obs

ll(null)

ll(model)

df

AIC

BIC

Crude
Adjusted

200
199

-1018.258
-1011.745

-993.7326
-963.3133

2
3

1991.465
1932.627

1998.062
1942.507

150
100

Missing sex

50

Blood pressure

200

250

weight10
Males
_cons

Crude

50
09-11

100
weight
H.S.

150
6

Common use

Pie and bar

Measure 3

Measure 1

Measure 2

mean of v1

09-11

H.S.

mean of v2

mean of v3

Bar-Dot-Line evolution

0
0

Parity

3+

mean of v2

(mean) v1

mean of v3

(mean) v2

3+

(mean) v3

mean of v1

1
(mean) v1

09-11

Parity
(mean) v2

H.S.

3+
(mean) v3

The workhorses:
Scatter and density

4000
3000
2000
1000

Birth weight

5000

Scatterplot

250

260

270
280
290
Gestational age

300

310

Density
Density

Boxplot

kdensity weight

2000

09-11

weight

4000

6000

graph hbox weight

2,000

H.S.

weight

4,000

6,000

12

Density with boxplot information

min

200

2040

25%

3180

Weight

50% 75%

3600

3940

w max

5350

5080

N=583
09-11

H.S.

13

Scatter and density plots


for many types of data
Y-type
Cont
Cont
Cont
Binary

X-type

Scatter

Cont
Cat
Cont

x
x
x

Density
x
x

x-normal use
x-suggested use

09-11

H.S.

14

Plots in analysis

Continuous outcome

Continuous by 1 category

min

200

2040

25%

3180

Weight

50% 75%

3600

3940

w max

5350

5080

N=583
09-11

H.S.

17

Continuous by 2 categories
Is birth weight the same for boys and girls?
Density plot

4000
3000
2000

Birth weight

5000

6000

Scatterplot

Boy

sex

Equal means? Linear effect?


Outliers?

Girl

2000

3000

4000
Birth weight

Equal variances?

5000

6000

Continuous by 3 categories
Is birth weight the same over parity?
Scatterplot

Density plot
0
1
2+

2000

Equal means? Linear effect?


Outliers?
09-11

3000

4000
Birth weight, g

5000

6000

Equal variances?
H.S.

19

Continuous by continuous
Does birth weight depend on gestational age?
Density plot

4000
3000
2000
1000

Birth weight

5000

Scatterplot

250

260

270
280
290
Gestational age

Equal means? Linear effect?


Outliers?
09-11

300

H.S.

310

20

Binary outcome

Binary by 2 categories

Jitter
and
line

.6
.6
.4
.4
0
0

.2
.2

Low birth weight

.8
.8

1
1

Does the low birth weight depend on sex?

Boy
Boy
Boy

sex
sex
sex

Girl
Girl

Binary by 3 categories

.2

.4

.6

.8

Does the low birth weight depend on parity?

1
Parity

2+

Binary by 3 categories, no scatter

.08
.04
0

Proportion low birth weight

.12

Does the low birth weight depend on parity?

1
Parity

2+

Scatter: binary by countinuous

1
.5
0
-.5

Proportion low birth weight

1.5

Does the low birth weight depend on gest.


age?

100

09-11

150

200
Gestational age

H.S.

250

300

25

Vous aimerez peut-être aussi