Analytic Hierarchy Process for Survey Data.

ahpsurvey

Overview

The ahpsurvey package provides a consistent methodology for researchers to reformat data and run the analytic hierarchy process (AHP), introduced by Thomas Saaty, on data that are formatted with the survey data entry mode. It is optimised for performing the AHP with many decision-makers, and provides tools and options for researchers to aggregate individual preferences and concurrently test multiple aggregation options. It also allows researchers to quantify, visualise and correct for inconsistent pairwise comparisons.

Installation

Install ahpsurvey directly from CRAN:

install.packages("ahpsurvey",repos = "http://cran.us.r-project.org")

Or, install the development version of ahpsurvey from Github with:

# install.packages("devtools")
devtools::install_github("frankiecho/ahpsurvey")

Usage

The ahpsurvey allows one to input a data.frame consisting of pairwise comparisons data collected through questionnaires and output an informative output of the aggregated priorities of all observations, the individual priorities, consistency ratios, and the most inconsistent pairwise comparisons.

library(ahpsurvey)
library(magrittr)

data(city200)
city200 %>% head()
#>   cult_fam cult_house cult_jobs cult_trans fam_house fam_jobs fam_trans
#> 1        2         -2         2         -6        -4       -4        -8
#> 2        2         -4         1         -4        -4       -2        -8
#> 3        4         -2         1         -3        -7       -3        -5
#> 4        8         -4         3         -4        -8        1        -7
#> 5        3         -3         5         -6        -8        1        -4
#> 6        6         -4         2         -4        -7       -2        -4
#>   house_jobs house_trans jobs_trans
#> 1          4          -3         -8
#> 2          4          -3         -7
#> 3          4          -3         -6
#> 4          4          -3         -9
#> 5          4          -3         -6
#> 6          4          -3         -6

Take a data.frame like that above and calculate the aggregated priorities of the 200 decision-makers.

## Define the attributes used
output <- ahp(city200, atts <- c("cult", "fam", "house", "jobs", "trans"), negconvert = TRUE, agg = TRUE)
#> [1] "Number of observations censored = 0"
output$aggpref
#>          AggPref  SD.AggPref
#> cult  0.15261018 0.033564038
#> fam   0.44827276 0.057695635
#> house 0.07052519 0.008844754
#> jobs  0.27579123 0.053734270
#> trans 0.03965027 0.006700507

And can show the detailed individual priorities of the 200 decision-makers and the consistency ratio of each decision-maker using that list:

head(output$indpref)[1:6]
#>        cult       fam      house      jobs      trans         CR
#> 1 0.1709466 0.4587181 0.08547330 0.2507636 0.03409845 0.06125366
#> 2 0.2291009 0.3935620 0.08292558 0.2531962 0.04121537 0.02962755
#> 3 0.1540045 0.4921905 0.08239372 0.2213908 0.05002052 0.06327989
#> 4 0.1242495 0.4634863 0.06162027 0.3159930 0.03465092 0.09308731
#> 5 0.1521676 0.3556904 0.07239889 0.3748108 0.04493236 0.10604443
#> 6 0.1536560 0.4738939 0.07106456 0.2516808 0.04970479 0.10740624

Further arguments allow you to specify the aggregation method, impute missing values and identify and correct inconsistent responses.

Functions

An overview of the functions in this package are as follows:

• ahp: A canned AHP routine
• ahp.mat: Generate AHP pairwise matrices from survey data
• ahp.indpref: Priority weights of individual decision-makers
• ahp.aggpref: Aggregate individual priorities (AIP)
• ahp.aggjudge: Aggregate individual judgements (AIJ)
• ahp.cr: Saaty’s Consistency Ratio
• ahp.error: The product between the pairwise comparison value and pj/pi
• ahp.pwerror: Finds the pairwise comparisons with the maximum amount of inconsistency
• ahp.missing: Impute missing pairwise comparsions
• ahp.harker: Replace inconsistent pairwise comparisons

Vignettes

For a detailed example of how the above function works, look no further than the vignettes, which are stored in /my-vignette.pdf. There, you can find a detailed step-by-step instruction of how to use the function using a simulated survey dataset and visualise the output using ggplot2.

Future development

I have plans to add the following features in the future, perhaps after I finish writing up my masters thesis :-(

• Multiple level of attributes: right now, you can always multiply the weights manually, but I’m looking to develop this feature in a convenient function
• Comparing alternatives: or a way to export the matrices to be used in other packages which does this
• Sensitivity analysis
• More ways to impute missing data
• Fuzzy AHP (or integration with existing packages)

Please let me know if there are any features which could be useful to you in a feature request or contribution.

Author

• Frankie Cho - Author & Maintainer - frankiecho

License

This project is licensed under the MIT License.