FECR - Use examples

Source: vignettes/FECR.Rmd

In this page we show of to perform FECR tests - Fecal Egg Count Reduction tests - from individual values of EPG: egg per gram of feces/fecal sediment.

Implemented methods refer to Cabaret & Berrag (2004), as well as our method that improve their proposed best approach.

We demonstrate the use of these methods on simulated data, as well as the toy example presented in this reference article.

Use example on simulated data

EPG simulation

We simulate EPG values (one per individual) by drawing from poisson distributions as follow:

library(FECR)

set.seed(42)

# Generate random EPG values, in the simple case of equal sample size
T1 <- rpois(50,100);T1.mean <- mean(T1)
T2 <- rpois(50,10);T2.mean <- mean(T2)
C1 <- rpois(50,100);C1.mean <- mean(C1)
C2 <- rpois(50,90);C2.mean <- mean(C2)

FECR calculations

When sample size is the same

The FECR() function of this FECR package takes vectors of EPG values as arguments T1, T2, C1, and C2.

From these values, you can specify a method between "Kochapakdee", "Dash", "Coles", "Cabaret1", "Cabaret2".

Note that some of these methods do not require all arguments: for these methods, you will need to specify what EPG values are inputted through named arguments (e.g. T2 = your.vector), otherwise an error message will be displayed.

The logical argument compute.CI allows for the computation of a confidence interval (95% by default) for individual-based methods (cf. Cabaret & Berrag, 2004)

# Calculation of FECR values according to different methods
FECR(T1.mean,T2.mean,method = "Kochapakdee")
## [1] 89.97253
FECR(T1,T2,C1,C2,method = "Dash")
## Warning in average_if_required(arg.list): a vector of epg has been provided in
## place of a mean. The arithmetic mean of the vector has been used.
## [1] 88.83641
FECR(T2 = T2.mean,C2 = C2.mean,method = "Coles")
## [1] 88.55799
FECR(T2 = T2,C2 = C2,method = "Coles")
## Warning in average_if_required(arg.list): a vector of epg has been provided in
## place of a mean. The arithmetic mean of the vector has been used.
## [1] 88.55799
FECR(T1,T2,method = "Cabaret1")
## [1] 89.89805
FECR(T1,T2,C1,C2,method = "Cabaret2")
## [1] 88.602
FECR(T1,T2,C1,C2,method = "Cabaret2",compute.CI = TRUE)
## [1] 88.602
## attr(,"CI")
##     2.5%    97.5% 
## 87.35018 89.75919
FECR(T1,T2,C1,C2,method = "MacIntosh1",compute.CI = TRUE)
## [1] 88.61238
## attr(,"CI")
##     2.5%    97.5% 
## 88.44445 88.76411

When sample size is uneven

The last included method, "MacIntosh1", is an improvement of the "Cabaret2" that:

  • depends on the arbitrary pair-matching of individuals in the control (C) and test (T) groups
  • cannot be applied to unequal sample-sizes without arbitrarily subset the larger group
# Generate random EPG values, in the case of unequal sample size
T1 <- rpois(60,100);T1.mean<- mean(T1)
T2 <- rpois(60,10);T2.mean<- mean(T2)
C1 <- rpois(50,100);C1.mean<- mean(C1)
C2 <- rpois(50,90);C2.mean<- mean(C2)

# Calculation of FECR values according to our sample-size unsensitive method
FECR(T1,T2,C1,C2,method = "Cabaret2",compute.CI = TRUE)
## Error in check_if_valid_parameters(arg.list): Some of T1,T2,C1,C2 are not of the same length.
FECR(T1,T2,C1,C2,method = "MacIntosh1")
## [1] 87.46709
FECR(T1,T2,C1,C2,method = "MacIntosh1",compute.CI = TRUE)
## [1] 87.46709
## attr(,"CI")
##     2.5%    97.5% 
## 87.28565 87.64213

The different aspects of this improvement are discussed further with demonstrations and examples here.