Link Search Menu Expand Document (external link)
How to Data

How to analyze the sample means of different treatment conditions (in R, using gplots and emmeans)

See all solutions.

Task

In a single-factor experiment with three or more treatment levels, how can we compare them to see which one impacts the outcome variable the most?

Solution

The solution below uses an example dataset about the teeth of 10 guinea pigs at three Vitamin C dosage levels (in mg) with two delivery methods (orange juice vs. ascorbic acid). (See how to quickly load some sample data.)

1

df <- ToothGrowth

To visually plot the means of the length of the tooth based on the Vitamin C dosage levels we can create a pointplot. We will use the gplots package. In the code below, bars=TRUE gives 95% confidence intervals for the means.

1
2
3

# install.packages("gplots") # If you have not yet installed it
library(gplots)
plotmeans(len~dose, data=df, bars=TRUE)

1
2
3
4
5
6

Attaching package: ‘gplots’


The following object is masked from ‘package:stats’:

    lowess

The point plot informs us that as the dosage levels increase, the tooth length also increases.

To obtain the actual numbers, we can use the code below. The first line converts the numerical dosage values to a categorical variable, which may not be necessary if your data was already categorical.

1
2
3

df$dose.factor = as.factor(df$dose)
aov1 = aov(len~dose.factor, data=df) 
model.tables(aov1, type='means')

1
2
3
4
5
6
7
8
9

Tables of means
Grand mean
         
18.81333 

 dose.factor 
dose.factor
   0.5      1      2 
10.605 19.735 26.100

If you wish to display the difference between the overall mean and the group means, you can simply omit the type='means' parameter.

1

model.tables(aov1)

1
2
3
4
5
6

Tables of effects

 dose.factor 
dose.factor
   0.5      1      2 
-8.208  0.922  7.287

To also see the specific values for the confidence intervals plotted earlier, we can use the emmeans package (Estimated Marginal Means or Least-Squares Means).

1
2
3

# install.packages("emmeans") # If you have not yet installed it
library(emmeans)
emmeans(aov1,'dose.factor')

1
2
3
4
5
6

 dose.factor emmean    SE df lower.CL upper.CL
 0.5           10.6 0.949 57     8.71     12.5
 1             19.7 0.949 57    17.84     21.6
 2             26.1 0.949 57    24.20     28.0

Confidence level used: 0.95

Content last modified on 24 July 2023.

See a problem? Tell us or edit the source.

Contributed by Krtin Juneja (KJUNEJA@falcon.bentley.edu)