In carlos-r-git/scriptsearch: A package for searching for text within existing scripts

The effect of different growth media on the growth of

three E.coli strains.

#

Introduction

The data in ecoli.txt are from an investigation of the growth of

three E.coli strains on four different media. The data are measures

of optical density (in arbitrary units) which gives an indication

of the number of cells in the medium. These data were analysed with

the aim of making recommendations to researchers about the best

strain or medium to use.

The data are organised into 3 columns

'data.frame': 96 obs. of 3 variables:

$ dens : num 2.9 9.8 12.8 7 11.6 10.9 14.4 5.1 17.7 12.1 ...

$ Strain: int 3 Strainds, 1, 2 and 3

$ medium: Factor w/ 4 levels "Circle", "Colibroth","Eplus", "GoCo"

Set up

working directory (in my case, yours will differ)

note: if you are using RStudio projects, this is unnecessary

setwd("M:/web/17C - 2018/scripts")

packages

library(tidyverse)

Hadley Wickham (2017). tidyverse: Easily Install and Load the 'Tidyverse'. R package version

1.2.1. https://CRAN.R-project.org/package=tidyverse

Import and tidy data

data are in ../data

ecoli <- read.table("../data/ecoli.txt", header = T)

check structure

str(ecoli)

'data.frame': 96 obs. of 3 variables:

$ dens : num 2.9 9.8 12.8 7 11.6 10.9 14.4

$ Strain: int 1 1 1 1 1 1 1 1 1 1 ...

$ medium: Factor w/ 4 levels "Circle","Colibroth"

names of the different media

levels(ecoli$medium)

"Circle" "Colibroth" "Eplus" "GoCo"

The data are in tidy format but Strain is treated as a measure.

Make $Strain a factor

ecoli$Strain <- factor(ecoli$Strain)

check structure

str(ecoli)

'data.frame': 96 obs. of 3 variables:

$ dens : num 2.9 9.8 12.8 7 11.6 10.9 14.4 5.1 17.7 12.

$ Strain: Factor w/ 3 levels "1","2","3": 1 1 1 1 1 1 1 1 1

$ medium: Factor w/ 4 levels "Circle","Colibroth",..

Exploratory Analysis

looking at the dataset there doesn't seem to be anything very

obviously non-normal (such as lots of values the same, too many

zeros, or extreme values

calculate some summary statistics

summary for strains

summarySE(ecoli, measurevar = "dens", groupvars = c("Strain"))

Strain N dens sd se ci

1 1 32 12.60625 5.216657 0.9221833 1.880805

2 2 32 14.94063 6.296177 1.1130174 2.270014

3 3 32 15.29063 6.289231 1.1117895 2.267510

Strain 3 has grown the best on average

summary for media

summarySE(ecoli, measurevar = "dens", groupvars = c("medium"))

medium N dens sd se ci

1 Circle 24 12.17917 6.440293 1.314619 2.719497

2 Colibroth 24 16.79583 5.072473 1.035414 2.141917

3 Eplus 24 15.92500 5.314480 1.084814 2.244108

4 GoCo 24 12.21667 5.951519 1.214849 2.513106

Colibroth promotes best growth on average but Eplus is close

Treatment combination summary. Summary will be useful for figure

ecolisummary <- summarySE(ecoli, measurevar = "dens", groupvars = c("Strain", "medium"))

Strain medium N dens sd se ci

1 1 Circle 8 9.3125 3.968604 1.403114 3.317836

2 1 Colibroth 8 14.6875 4.837779 1.710413 4.044485

3 1 Eplus 8 17.3125 3.267781 1.155335 2.731933

4 1 GoCo 8 9.1125 3.751357 1.326305 3.136213

5 2 Circle 8 10.6000 7.297945 2.580213 6.101235

6 2 Colibroth 8 15.3125 4.888014 1.728174 4.086482

7 2 Eplus 8 15.7250 6.790697 2.400874 5.677165

8 2 GoCo 8 18.1250 4.278768 1.512773 3.577140

9 3 Circle 8 16.6250 5.695800 2.013770 4.761808

10 3 Colibroth 8 20.3875 3.860954 1.365053 3.227838

11 3 Eplus 8 14.7375 5.670207 2.004721 4.740412

12 3 GoCo 8 9.4125 4.909884 1.735906 4.104766

it looks as though different media favour different strains.

conclusion: n = 8 for each treatment combination is quite

small but the design is balanced, the data have decimal places,

lack repeated repeated values and extreme values

We would expect such OD data to be normally distributed therefore

two-way ANOVA is the desired test if the assumptions are met

Statistical Analysis

The assumptions of the two-way ANOVA test are 1. the residuals are

normally distributed and 2. the variances is homogenous.

These will be investigated in the ANOVA model constructed.

Carrying out an ANOVA to examine the effect of strain, the effect

of medium and whether their effects are independent

mod <- aov(data = ecoli, dens ~ Strain * medium) summary(mod)

Medium appears significant (F = 5.5, d.f. = 3. 84, p = 0.0017);

strain appears NS but there is a signifcant interaction present

indicating that the effect of medium depends on the strain

that is being considered.

investigate the dependence of effects with an interactions plot

consider both ways round

interaction.plot(ecoli$Strain, ecoli$medium,ecoli$dens) interaction.plot(ecoli$medium, ecoli$Strain, ecoli$dens)

No single medium is best for all strains, GoCo is best

for strain 2 but worse for strains 1 and 3. If you wanted

to chose one medium which was 'ok' for any strain, you

might chose Colibroth or Eplus as the other two media are e

specially bad for at least one of the strains. Eplus is the

most independent of strain. Post-hoc tests are required to

establish is these patterns are signifcant

Carrying out a Tukey Honest Signifcant differences test

to see where differences bewteen treatment groups. There are many

comparisons and understanding the results means going through

these carefully. There are no shortcuts.

There are many comparisons so a plot is not that useful here

TukeyHSD(mod)

Significant comparisons listed only

$Strain:medium

diff lwr upr p adj

3:Colibroth-1:Circle 11.0750 2.543727 19.606273 0.0020260

2:GoCo-1:Circle 8.8125 0.281227 17.343773 0.0365730

3:Colibroth-2:Circle 9.7875 1.256227 18.318773 0.0113418

1:GoCo-3:Colibroth -11.2750 -19.806273 -2.743727 0.0015266

2:GoCo-1:GoCo 9.0125 0.481227 17.543773 0.0290721

3:GoCo-2:GoCo -8.7125 -17.243773 -0.181227 0.0409305

Notable from this and summary data:

If you are using GoCo, use strain 2 (it's better than 1 and

3 on GoCo). If you're only using Circle use Strain 3. But

best combination is Colibroth and Strain 3. No strain

goes especially badly on Colibroth or Eplus

Figure

uses the summary data

ggplot(ecolisummary, aes(x = medium, y = dens, fill=Strain) ) + geom_bar(stat="identity", colour="black", position = position_dodge()) + geom_errorbar(aes(ymin = dens - se, ymax = dens + se), width = 0.4, position = position_dodge(0.9)) + scale_fill_manual(values = c("#FFFFFF", "#CCCCCC", "#999999")) + ylab("Optical density (units)") + ylim(0, 30) + xlab("Medium") + theme(panel.background = element_rect(fill = "white"))+ theme(axis.line.x = element_line(color = "black"), axis.line.y = element_line(color = "black")) + # 2:GoCo-1:GoCo annotate("segment", x = 3.7, xend = 3.95, y = 20.5, yend = 20.5, colour = "black") + annotate("segment", x = 3.7, xend = 3.7, y = 20.5, yend = 20, colour = "black") + annotate("segment", x = 3.95, xend = 3.95, y = 20.5, yend = 20, colour = "black") + annotate("text", x = 3.85, y = 20.8, label = "", size = 5) + # 3:GoCo-2:GoCo annotate("segment", x = 4.3, xend = 4.05, y = 20.5, yend = 20.5, colour = "black") + annotate("segment", x = 4.3, xend = 4.3, y = 20.5, yend = 20, colour = "black") + annotate("segment", x = 4.05, xend = 4.05, y = 20.5, yend = 20, colour = "black") + annotate("text", x = 4.15, y = 20.8, label = "", size = 5)+ # 1:GoCo-3:Colibroth
annotate("segment", x = 3.7, xend = 2.35, y = 22.5, yend = 22.5, colour = "black") + annotate("segment", x = 2.35, xend = 2.35, y = 22.5, yend = 22, colour = "black") + annotate("segment", x = 3.7, xend = 3.7, y = 22.5, yend = 22, colour = "black") + annotate("text", x = 3, y = 22.8, label = "", size = 5) + # 2:GoCo-3:Colibroth annotate("segment", x = 4, xend = 2.35, y = 23.5, yend = 23.5, colour = "black") + annotate("segment", x = 2.35, xend = 2.35, y = 23.5, yend = 23, colour = "black") + annotate("segment", x = 4, xend = 4, y = 23.5, yend = 23, colour = "black") + annotate("text", x = 3.1, y = 23.8, label = "", size = 5) + # 3:Colibroth-2:Circle annotate("segment", x = 1, xend = 2.25, y = 22.5, yend = 22.5, colour = "black") + annotate("segment", x = 2.25, xend = 2.25, y = 22.5, yend = 22, colour = "black") + annotate("segment", x = 1, xend = 1, y = 22.5, yend = 22, colour = "black") + annotate("text", x = 1.6, y = 22.8, label = "*", size = 5) + # 3:Colibroth-1:Circle annotate("segment", x = 0.7, xend = 2.25, y = 23.5, yend = 23.5, colour = "black") + annotate("segment", x = 2.25, xend = 2.25, y = 23.5, yend = 23, colour = "black") + annotate("segment", x = 0.7, xend = 0.7, y = 23.5, yend = 23, colour = "black") + annotate("text", x = 1.5, y = 23.8, label = "", size = 5)+ # 2:GoCo-1:Circle annotate("segment", x = 0.7, xend = 4, y = 24.5, yend = 24.5, colour = "black") + annotate("segment", x = 4, xend = 4, y = 24.5, yend = 24, colour = "black") + annotate("segment", x = 0.7, xend = 0.7, y = 24.5, yend = 24, colour = "black") + annotate("text", x = 2.4, y = 24.8, label = "*", size = 5)

Figure 2. The effect of media on the growth of different

strains of E.coli. Error bars are $\pm$ 1 S.E. Comparisons

significant under post-hoc testing with Tukey's Honest

Significant Difference test

carlos-r-git/scriptsearch documentation built on Sept. 1, 2020, 6:38 p.m.