Lecture 3. planning data analysis

Workshop on Analysis of Clinical Studies – Can Tho University of Medicine and Pharmacy – April 2012
Planning Data Analysis
Tuan V. Nguyen
Professor and NHMRC Senior Research Fellow
Garvan Institute of Medical Research
University of New South Wales
Sydney, Australia

Plan of analysis
•  Checking data consistency
•  Aims of analysis
•  Set up tables and graphs
•  Determine appropriate R packages
•  Doing analysis and documentation

Checking data consistency

Data integrity is VERY important
•  Data accuracy and integrity: extremely important
•  Data preparation: 90%, analysis: 10%

Main Steps for Data Preparation
•  Create the data file.
–  Original data
–  Interim data
–  Documentation
•  Clean the data
•  Process the data
•  Create an analysis-ready copy of the data
•  Document the data

Some critical points to format your data set
•  Do not include header, trailer information,
subtotals, or other extraneous information. For
descriptive purposes you may include one row
giving variable names.

Data layout
•  All analysis software accept data in table format (data
matrix)
–  For survey data, column usually represents single variable
(a single question), row represents an individual case
–  Different processing for multiple-response questions
Id Age Gender Service employed
Case 1 1 27 1 2 1
Case 2 2 19 2 1 2
Case 3 3 24 2 3 1

Format data so that each variable is in its own single column
For example,
and better not
and absolutely not

Columns must be explicitly filled with data
and not

•  All columns must have the same number of
rows and Missing data (empty cells) Use a blank
space or a . to indicate missing data.
- Missing value cannot be treated as “0”.
•  Keep an identical field, such the subject ID.

•  If there are multiple data files, do not rely on the file names
to carry variable information. For example, if separate files
are used for the results of two treatments, include a column
in each file containing the name of the treatment

Not like these !

Counted proportion data
If data consists of counted proportions,
e.g. number of individuals responding out of total number
of individuals,
do not reduce the data to percentages or proportions
beforehand.
It is recommended that both numerator and denominator
of the proportion be entered as separate columns.

For example,
Not
It is easy to compute proportions during the analysis if they
are required, but alternative analyses such as logistic
regression may be precluded if original counts are
unavailable.

•  Polytomous data.
If data consists of numbers falling into a number of mutually
exclusive classes, do not reduce to proportions or percentages
beforehand, but enter the integer counts.
For example,
not

•  Each column has its own criteria or “meaning”.
Narrow the definition for each variable means to create a
new variable.
For example,
gender, age → >45 year-old male

Checking distribution
•  Use histogram to check for distribution
–  Are they normally distributed
•  Use scatterplot to check for abnormal data points

Histogram of menarche
menarche
Frequency
0 20 40 60 80 100
0
100
200
300
400
Checking for outliers

Aims of analysis

Common aims of data analysis
•  Determine your aims!
–  What do you want to do ?
•  Description of single variable
•  Comparing between groups
•  Finding association between variables
•  Development of predictive models

Descriptive analysis of single variable
•  Continuous variables
–  Looking for distribution (hist, boxplot, plot)
–  Mean, SD, median, interquartile range (psych package:
describe, describe.by)
•  Categorical variables
–  Proportion (table, gmodels: CrossTable)
–  Counts
–  95% confidence interval

Comparison between groups
–  T-test, ANOVA (t.test, anova, lm)
–  Z-test, Chi-squared analysis (Cross.Table,
chisq.test, fisher.test)
–  Binomial test (prop.test)

Finding association between variables
–  Simple linear regression (lm)
–  Multiple linear regression (lm)
–  Logistic regression model (glm)
–  Binomial regression (glm)
–  Cox’s proportional regression model (survfit,
survdiff, coxph)

An example

Consider the vitamin D study
•  Cross-sectional study
•  N = 558 participants, 222 men + 336 women, aged 13
– 83 yr
•  Aims:
–  Prevalence of vitamin D deficiency in Vietnamese
–  Risk factors for vitamin D deficiency
–  Determinants of vitamin D

Variables in dataset
Basic demographic and clinical
data
•  id
•  sex
•  dob
•  address
•  region
•  age
•  menarche
•  height
•  weight
•  bmi
•  fracture
•  alcohol, coffee, tea
•  season
Biochemical and BMD data
•  estradiol
•  testo
•  vitd
•  pth
•  xlap
•  fnbmd
•  hipbmd
•  lsbmd

Data in Excel (.csv format)

Plan of analysis
•  Checking data consistency
•  Checking the distribution of vitamin D and hormones
•  Descriptive statistics of all variables for men and
women separately

Plan of analysis
•  Define vitamin insufficiency, deficiency – call this
def
•  Find the frequency of def by:
–  sex, age group, urban vs rural, season
•  Find risk factors for def
•  Find the correlation between vitamin D and PTH. Is
there a threshold whereby vitamin D is stable

Statistical methods
Checking data consistency Graphical analysis
Checking the distribution of vitamin D and
hormones
Histogram
Descriptive statistics of all variables for men
and women separately
T-test for continuous
variables and z test
for categorical
variables
Prevalence of vit D deficiency for men and
women, and various factors
Counting, Chi square
test
Find risk factors for def Logistic regression
Find the correlation between vitamin D and
PTH. Is there a threshold whereby vitamin D is
stable
Linear regression

What we want to have …
Table 1. Characteristics of participants
Variable Men Women P-value
Age
Height
Weight
BMI
25(OH)D
PTH
Etc.

Table 2. Prevalence of vitamin D deficiency
Risk factor N Prevalence 95% CI
Sex
Men
Women
Age
<20
20-29
30-39
40-49
50-59
60+
Season
Winter
Non-winter
Residence
Rural
Urban

Table 3. Risk factors for vitamin D deficiency
Risk factor Odds ratio 95% CI P-value
Age
Sex
BMI
Season
Residence
Etc.
Which determinants are statistically significant ?
How much variance of vit D deficiency can be explained by these risk
factors?
Which one is important?

Table 4. Determinants of 25(OH) levels
Determinant Unit Regression
coefficient
P-value
Age
Sex
BMI
Season
Residence
Etc.
Which determinants are statistically significant ?
How much variance of vit D can be explained by these determinants?

PTH
Vit D
Age
Vit D
Men
Women
?

Winter
Vit D
Men
Vit D
Summer Women
Spring Autumn

Procedures of analysis
•  Arrange data in Excel
•  Use R packages to analysis
–  pysch (for descriptive analyses)
–  rms (for logistic regression analysis)
–  Hmisc (for data manipulation)
–  xtab (for contingency tables)

Read-in data
# Tell R where the data are stored
setwd("C:/Documents and Settings/Tuan/My Documents/_Current Projects/
_Vietnam/Huong/Vitamin D")
# Reading the data into a dataset called vd
vd = read.csv("vitaminD.csv", header=T, na.strings=" ")
# What variables are in vd
names(vd)
# Tell R that we will work on vd from now on
attach(vd)

Checking data consistency
# Let’s get some basic statistics for all variables
# First, we load the package psych
library(psych)
# then get statistics
describe(vd)

> describe(vd, skew=FALSE)
var n mean sd median trimmed mad min max range se
id* 1 558 278.37 159.85 279.50 278.58 204.60 1.00 554.00 553.00 6.77
sex 2 558 1.60 0.49 2.00 1.63 0.00 1.00 2.00 1.00 0.02
address* 3 558 4.65 2.17 4.50 4.68 3.71 1.00 8.00 7.00 0.09
region 4 558 2.60 1.30 2.50 2.63 2.22 1.00 4.00 3.00 0.05
age 5 558 46.63 17.79 49.00 46.82 20.76 13.00 83.00 70.00 0.75
menarche 6 552 32.61 33.04 17.00 26.84 2.97 0.00 99.00 99.00 1.41
estradiol 7 558 67.57 74.74 41.35 51.03 25.69 5.00 476.10 471.10 3.16
testo 8 558 2.67 3.26 0.50 2.19 0.68 0.02 13.85 13.83 0.14
vitd 9 558 25.69 8.62 25.83 25.64 7.31 4.00 59.87 55.87 0.36
xlap 10 244 0.39 0.27 0.31 0.35 0.21 0.02 1.57 1.55 0.02
pth 11 558 30.62 12.65 29.04 29.52 11.79 1.20 93.53 92.33 0.54
height 12 558 157.71 7.92 157.00 157.60 7.41 134.00 181.00 47.00 0.34
weight 13 558 50.70 8.08 50.00 50.23 7.41 32.00 85.00 53.00 0.34
bmi 14 558 20.33 2.63 20.00 20.18 2.97 14.00 31.00 17.00 0.11
fnbmd 15 558 0.72 0.12 0.71 0.72 0.12 0.39 1.11 0.72 0.01
hipbmd 16 558 0.81 0.12 0.81 0.81 0.11 0.29 1.17 0.88 0.01
lsbmd 17 554 0.86 0.15 0.87 0.86 0.15 0.42 1.48 1.05 0.01
fracture 18 558 4.90 16.70 2.00 2.00 0.00 1.00 99.00 98.00 0.71
alcohol 19 557 2.63 9.19 2.00 1.82 0.00 1.00 99.00 98.00 0.39
coffee 20 552 1.87 0.34 2.00 1.96 0.00 1.00 2.00 1.00 0.01
tea 21 552 1.61 0.49 2.00 1.64 0.00 1.00 2.00 1.00 0.02
ruralurban 22 558 1.69 0.46 2.00 1.74 0.00 1.00 2.00 1.00 0.02
location 23 558 0.31 0.46 0.00 0.26 0.00 0.00 1.00 1.00 0.02
season 24 558 2.18 0.81 2.00 2.22 1.48 1.00 3.00 2.00 0.03

Histogram of menarche
menarche
Frequency
0 20 40 60 80 100
0
100
200
300
400
> hist(menarche)

Descriptive analysis by sex
describe.by(vd, sex, skew=F, interp=FALSE)
group: 1
id* 1 222 296.79 166.79 302.50 301.37 197.19 3.00 554.00 551.00 11.19
sex 2 222 1.00 0.00 1.00 1.00 0.00 1.00 1.00 0.00 0.00
address* 3 222 4.85 2.06 5.00 4.94 2.97 2.00 7.00 5.00 0.14
region 4 222 2.72 1.25 3.00 2.77 1.48 1.00 4.00 3.00 0.08
age 5 222 46.67 19.30 49.00 46.57 25.20 14.00 83.00 69.00 1.30
menarche 6 220 53.99 40.85 20.00 53.31 7.41 12.00 99.00 87.00 2.75
estradiol 7 222 42.76 12.46 41.33 42.19 10.90 8.95 92.33 83.38 0.84
testo 8 222 6.24 2.25 6.05 6.16 2.16 0.84 13.85 13.01 0.15
vitd 9 222 28.57 8.94 28.29 28.45 8.46 4.00 59.87 55.87 0.60
xlap 10 100 0.45 0.31 0.35 0.41 0.26 0.02 1.57 1.55 0.03
pth 11 222 31.76 13.50 29.69 30.29 12.34 8.63 93.53 84.90 0.91
height 12 222 164.20 6.03 164.00 164.07 5.93 148.00 181.00 33.00 0.40
weight 13 222 54.46 8.44 53.00 54.06 7.41 35.00 85.00 50.00 0.57
bmi 14 222 20.13 2.72 20.00 19.96 2.97 15.00 31.00 16.00 0.18
fnbmd 15 222 0.76 0.12 0.76 0.75 0.12 0.39 1.11 0.72 0.01
hipbmd 16 222 0.85 0.12 0.84 0.85 0.12 0.29 1.17 0.88 0.01
lsbmd 17 220 0.90 0.14 0.91 0.90 0.14 0.53 1.48 0.94 0.01
fracture 18 222 4.14 14.43 2.00 2.00 0.00 1.00 99.00 98.00 0.97
alcohol 19 222 1.88 6.57 1.00 1.43 0.00 1.00 99.00 98.00 0.44
coffee 20 221 1.79 0.41 2.00 1.86 0.00 1.00 2.00 1.00 0.03

group: 2
id* 1 336 266.19 154.14 266.00 264.52 202.37 1.00 551.00 550.00 8.41
sex 2 336 2.00 0.00 2.00 2.00 0.00 2.00 2.00 0.00 0.00
address* 3 336 4.51 2.23 4.00 4.51 2.97 1.00 8.00 7.00 0.12
region 4 336 2.53 1.33 2.00 2.53 1.48 1.00 4.00 3.00 0.07
age 5 336 46.60 16.74 49.00 47.04 19.27 13.00 82.00 69.00 0.91
menarche 6 332 18.44 14.42 16.00 16.00 2.97 0.00 99.00 99.00 0.79
estradiol 7 336 83.96 92.24 41.74 65.28 36.46 5.00 476.10 471.10 5.03
testo 8 336 0.32 0.59 0.20 0.23 0.19 0.02 7.91 7.89 0.03
vitd 9 336 23.79 7.86 24.26 23.93 7.10 4.00 59.60 55.60 0.43
xlap 10 144 0.34 0.24 0.28 0.31 0.19 0.06 1.26 1.20 0.02
pth 11 336 29.87 12.03 28.56 29.03 11.32 1.20 92.87 91.67 0.66
height 12 336 153.42 5.82 154.00 153.53 5.93 134.00 168.00 34.00 0.32
weight 13 336 48.21 6.78 48.00 47.94 7.41 32.00 74.00 42.00 0.37
bmi 14 336 20.46 2.56 20.00 20.33 2.97 14.00 30.00 16.00 0.14
fnbmd 15 336 0.70 0.11 0.69 0.70 0.12 0.39 1.00 0.60 0.01
hipbmd 16 336 0.78 0.12 0.79 0.78 0.11 0.48 1.10 0.62 0.01
lsbmd 17 334 0.83 0.15 0.85 0.84 0.16 0.42 1.22 0.80 0.01
fracture 18 336 5.40 18.04 2.00 2.00 0.00 1.00 99.00 98.00 0.98
alcohol 19 335 3.12 10.56 2.00 2.00 0.00 1.00 99.00 98.00 0.58
coffee 20 331 1.92 0.28 2.00 2.00 0.00 1.00 2.00 1.00 0.02
tea 21 331 1.80 0.40 2.00 1.87 0.00 1.00 2.00 1.00 0.02
ruralurban 22 336 1.65 0.48 2.00 1.69 0.00 1.00 2.00 1.00 0.03
location 23 336 0.35 0.48 0.00 0.31 0.00 0.00 1.00 1.00 0.03
season 24 336 2.16 0.85 2.00 2.20 1.48 1.00 3.00 2.00 0.05

Distribution of some continuous data
Histogram of vitd
vitd
Frequency
0 10 20 30 40 50 60
0
50
150
Histogram of pth
pth
Frequency
0 20 40 60 80 100
0
100
200
Histogram of bmi
bmi
Frequency
15 20 25 30
0
50
150
Histogram of age
age
Frequency
20 40 60 80
0
40
par(mfrow=c(2,2))
hist(vitd); hist(pth); hist(bmi); hist(age)

Defining vitamin D insufficiency
library(Hmisc)
insuff = cut2(vitd, 30)
def = vitd
def = replace(def, vitd < 20.0, 1)
def = replace(def, vitd >= 20, 0)
library(gmodels)
CrossTable(def, sex, digits=3, chisq=T, fisher=T)
CrossTable(def, season, digits=3, chisq=T, fisher=T)
CrossTable(def, location, digits=3, chisq=T, fisher=T)

Total Observations in Table: 558
| def
sex | 0 | 1 | Row Total |
-------------|-----------|-----------|-----------|
1 | 187 | 35 | 222 |
| 1.988 | 6.290 | |
| 0.842 | 0.158 | 0.398 |
| 0.441 | 0.261 | |
| 0.335 | 0.063 | |
-------------|-----------|-----------|-----------|
2 | 237 | 99 | 336 |
| 1.313 | 4.156 | |
| 0.705 | 0.295 | 0.602 |
| 0.559 | 0.739 | |
| 0.425 | 0.177 | |
-------------|-----------|-----------|-----------|
Column Total | 424 | 134 | 558 |
| 0.760 | 0.240 | |
-------------|-----------|-----------|-----------|
Vitamin deficiency by sex

Statistics for All Table Factors
Pearson's Chi-squared test
------------------------------------------------------------
Chi^2 = 13.74685 d.f. = 1 p = 0.0002091717
Pearson's Chi-squared test with Yates' continuity correction
------------------------------------------------------------
Chi^2 = 13.00639 d.f. = 1 p = 0.0003104303
Fisher's Exact Test for Count Data
------------------------------------------------------------
Sample estimate odds ratio: 2.22876
Alternative hypothesis: true odds ratio is not equal to 1
p = 0.0001784270
95% confidence interval: 1.427180 3.540137
Vitamin deficiency by sex: test

| def
season | 0 | 1 | Row Total |
-------------|-----------|-----------|-----------|
1 | 77 | 64 | 141 |
| 8.479 | 26.828 | |
| 0.546 | 0.454 | 0.253 |
| 0.182 | 0.478 | |
| 0.138 | 0.115 | |
-------------|-----------|-----------|-----------|
2 | 147 | 30 | 177 |
| 1.163 | 3.679 | |
| 0.831 | 0.169 | 0.317 |
| 0.347 | 0.224 | |
| 0.263 | 0.054 | |
-------------|-----------|-----------|-----------|
3 | 200 | 40 | 240 |
| 1.705 | 5.396 | |
| 0.833 | 0.167 | 0.430 |
| 0.472 | 0.299 | |
| 0.358 | 0.072 | |
-------------|-----------|-----------|-----------|
Column Total | 424 | 134 | 558 |
| 0.760 | 0.240 | |
-------------|-----------|-----------|-----------|
Vitamin deficiency by season

Risk factors for vitamin D deficiency: univariate analysis
library(epicalc)
logistic.display(glm(def ~ sex, family=binomial))
logistic.display(glm(def ~ location, family=binomial))
OR(95%CI) P(Wald's test) P(LR-test)
sex (cont. var.) 2.23 (1.45,3.43) < 0.001 < 0.001
Log-likelihood = -300.4421
No. of observations = 558
AIC value = 604.8842
location: 1 vs 0 3.71 (2.47,5.58) < 0.001 < 0.001

Risk factors for vitamin D deficiency: univariate analysis
logistic.display(glm(def ~ bmi, family=binomial))
logistic.display(glm(def ~ factor(season), family=binomial))
bmi (cont. var.) 1.05 (0.98,1.13) 0.174 0.176
OR lower95ci upper95ci Pr(>|Z|)
factor(season)2 0.2455357 0.1468710 0.4104812 8.506640e-08
factor(season)3 0.2406250 0.1497158 0.3867354 4.002872e-09

Risk factors for vitamin D deficiency: multivariable
analysis
OR lower95ci upper95ci Pr(>|Z|)
age 0.9821236 0.9698786 0.9945232 0.004833844
sex 1.9543516 1.2394551 3.0815883 0.003927809
bmi 0.9878399 0.9038924 1.0795839 0.787155190
location 2.4925512 0.9452428 6.5727150 0.064872828
factor(season)2 0.5482407 0.2237126 1.3435448 0.188766835
factor(season)3 0.5817646 0.1990562 1.7002735 0.322202383
logistic.display(glm(def ~ age+sex+bmi+location+factor(season),
family=binomial))

Risk factors for vitamin D deficiency: binomial regression
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -2.060883 0.411777 -5.005 5.59e-07 ***
age -0.013404 0.004877 -2.748 0.0060 **
sex 0.503312 0.198337 2.538 0.0112 *
location 0.938540 0.175347 5.352 8.68e-08 ***
library(MASS)
summary(glm.nb(def ~ age+sex+location, data=vd))

Determinants of vitamin D: multivariable analysis
library(leaps)
library(car)
library(MASS)
library(relaimpo)
# stepwise regression
fit = lm(vitd~age+bmi+sex+factor(season), data=vd)
step = stepAIC(fit, direction="both")
step$anova
# All Subsets Regression
fit2 = regsubsets(vitd~age+bmi+sex+factor(season),data=vd,nbest=5)
summary(fit2)
plot(fit2,scale="r2")
subsets(fit2, statistic="rsq")

Determinants of vitamin D : multivariable analysis
# Calculate Relative Importance for Each Predictor
library(relaimpo)
fit = lm(vitd~age+bmi+sex+factor(season), data=vd)
calc.relimp(fit, rela=TRUE)
# Bootstrap Measures of Relative Importance (1000
samples)
library(boot)
boot = boot.relimp(fit, b = 1000), rank = TRUE,
diff=TRUE, rela=TRUE)
booteval.relimp(boot)
# print result
plot(booteval.relimp(boot,sort=TRUE)) # plot result

Relative importance metrics:
lmg
factor(season) 0.48695621
age 0.06065485
bmi 0.02030591
sex 0.43208303
Average coefficients for different model sizes:
1group 2groups 3groups 4groups
age 0.04504289 0.05004166 0.04961454 0.04470275
bmi -0.23392346 -0.14266033 -0.05454423 0.03080951
sex -4.77558719 -4.61454708 -4.46527769 -4.34587273
`factor(season)`2 5.45826902 5.18184506 4.86820278 4.52678038
`factor(season)`3 5.93707270 5.90083953 5.81691391 5.69611558

Determinants of vitamin D: stepwise regression
> fit = lm(vitd~age+bmi+sex+factor(season), data=vd)
> step = stepAIC(fit, direction="both")
Start: AIC=2322.88
vitd ~ age + bmi + sex + factor(season)
Df Sum of Sq RSS AIC
- bmi 1 3 35095 2321
<none> 35092 2323
- age 1 324 35416 2326
- sex 1 2425 37517 2358
- factor(season) 2 2616 37708 2359
Step: AIC=2320.93
vitd ~ age + sex + factor(season)
Df Sum of Sq RSS AIC
<none> 35095 2321
+ bmi 1 3 35092 2323
- age 1 371 35466 2325
- sex 1 2422 37517 2356
- factor(season) 2 2895 37989 2361

Determinants of vitamin D: stepwise regression
> step$anova
Stepwise Model Path
Analysis of Deviance Table
Initial Model:
vitd ~ age + bmi + sex + factor(season)
Final Model:
vitd ~ age + sex + factor(season)
Step Df Deviance Resid. Df Resid. Dev AIC
1 552 35091.86 2322.882
2 - bmi 1 2.975332 553 35094.83 2320.929

Determinants of vitamin D: “final model”
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 26.66091 1.64606 16.197 < 2e-16 ***
age 0.04591 0.01899 2.418 0.0159 *
sex -4.33997 0.70246 -6.178 1.26e-09 ***
factor(season)2 4.47858 0.91536 4.893 1.31e-06 ***
factor(season)3 5.62941 0.84679 6.648 7.14e-11 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Residual standard error: 7.966 on 553 degrees of freedom
Multiple R-squared: 0.1521, Adjusted R-squared: 0.146
F-statistic: 24.81 on 4 and 553 DF, p-value: < 2.2e-16
fit = lm(vitd ~ age+sex+factor(season), data=vd)
summary(fit)

Relationship between vitamin D and PTH
library(car)
scatterplot(vitd ~ pth | sex, pch=c(1, 15),
xlab="PTH", ylab="25(OH)D")
0 20 40 60 80
10
20
30
40
50
60
PTH
25(OH)D

> fit = lm(vitd ~ pth, data=vd)
> summary(fit)
Coefficients:
(Intercept) 26.70950 0.95613 27.935 <2e-16 ***
pth -0.03329 0.02886 -1.153 0.249
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
F-statistic: 1.33 on 1 and 556 DF, p-value: 0.2492

> fit = lm(vitd ~ pth+sex+pth:sex, data=vd)
> summary(fit)
Coefficients:
(Intercept) 38.82787 3.09515 12.545 < 2e-16 ***
pth -0.17214 0.09074 -1.897 0.0583 .
sex -7.36461 1.86973 -3.939 9.23e-05 ***
pth:sex 0.08091 0.05586 1.448 0.1481
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
F-statistic: 16.47 on 3 and 554 DF, p-value: 2.907e-10

Let’s have an overall picture
library(psych)
dd = cbind(age, bmi, vitd, pth)
pairs.panels(dd)
age
15 20 25 30
0.27 0.09
0 20 40 60 80
20
40
60
80
0.10
15
20
25
30
bmi
-0.07 0.02
vitd
10
30
50
-0.05
20 40 60 80
0
20
40
60
80
10 30 50
pth

Error plot
error.bars.by(vitd, sex, bars=TRUE, labels=c("Men",
"Women"), ylab = "25(OH)D", xlab=" ", ylim=c(0,30))
Men Women
0.95% confidence limits
25(OH)D
0
10
20
30
40
50
60
0
10
20
30
40
50
60

Lecture 3. planning data analysis

More Related Content

What's hot (20)

Similar to Lecture 3. planning data analysis (20)

More from Hau Pham (12)

Recently uploaded (20)

Lecture 3. planning data analysis