indirectCalibration function

Function performing the indirect calibration

indirectCalibration is a function for the indirect calibration procedure as described by Ragin (2008). It uses a binomial or a beta regression for tranforming raw scores into calibrated scores. In our opinion, using a fractional polynomial may not be appropriate to this case. In fact, we do not deal with proportions. This function requires the package betareg.

indirectCalibration(x, x_cal, binom = TRUE)

Arguments

• x: vector of raw scores.
• x_cal: vector of theoretically calibrated scores.
• binom: logical. If indirect calibration has to be performed using binomial regression or beta regression. The default is TRUE, which means that binomial regression is used.

Returns

It returns a vector of indirectly calibrated values.

References

Ragin, C. C. (2008) Redesigning Social Inquiry: Fuzzy Sets and Beyond, The Chicago University Press: Chicago and London.

Schneider, C. Q., Wagemann, C. (2012) Set-Theoretic Methods for the Social Sciences, Cambridge University Press: Cambridge.

Author(s)

Mario Quaranta

Examples

# Generate fake data
set.seed(4)
x <- runif(20, 0, 1)

# Find quantiles
quant <- quantile(x, c(.2, .4, .5, .6, .8))

# Theoretical calibration
x_cal <- NA  
x_cal[x <= quant[1]] <- 0
x_cal[x > quant[1] & x <= quant[2]] <- .2
x_cal[x > quant[2] & x <= quant[3]] <- .4
x_cal[x > quant[3] & x <= quant[4]] <- .6
x_cal[x > quant[4] & x <= quant[5]] <- .8
x_cal[x > quant[5]] <- 1
x_cal

# Indirect calibration (binomial)
a <- indirectCalibration(x, x_cal, binom = TRUE)

# Indirect calibration (beta regression)
b <- indirectCalibration(x, x_cal, binom = FALSE)

# Correlation
cor(a, b)

# Plot
plot(x, a); points(x, b, col = "red")