In mabafaba/hypegrammaR: A grammar of hypothesis test driven analysis

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  eval = FALSE
)

Main Steps

once: setup github repository
once: install packages
load packages
load inputs from csv files
preprocess inputs
make sure all inputs can be matched
create analysisplan
run analysisplan
export results

Details

DEFINE YOUR ANALYSIS

create analysisplan
create recoding plan
add newly created variables to kobo tool - to pretend they had been collected

LOAD PACKAGES

library(koboquest)
library(hypegrammaR)
library(composr)
library(kobostandards)
library(dplyr)
library(xlsformfill)

PREPARE INPUTS

load inputs as data.frames from csv
- questionnaire questions sheet ("questions") (excel -> save active sheet as csv)
- questionnaire choices sheet ("choices")
  - add external choices to the regular choices sheet
  - replace "select_one_external" and "select_multiple_external" by "select_one"
- analysisplan
- recoding plan
- data
- sampling frame
(Generate data)
create koboquest questionnaire object

mydata <- read.csv("./input/data.csv", stringsAsFactors = F)
...

mydata <- xlsform_fill(questions,choices,1000)

load_questionnaire(mydata, questions, choices)

RECODING

batch recode from recodingplan

SAMPLING STRATEGY PREPARATIONS

stratification weighting:
- create a data variable whos values match exactly the strata names in the sampling frame
  - for all stratification elements: districts & population groups!
(cluster weighting: maybe.. depending on sampling strategy)
- same as stratitifaction
- use surveyweights::combine_weighting_functions() to combine weighting functions
Create unique cluster id variable in the data
create weighting function (map_to_weighting) -> weighting

samplingframe_tidy<- samplingframe_tidy %>%
  mutate(stratum_id = paste0(district.pcode,"_",population_group))


data <- data %>%
  mutate(data_stratum_id = paste0("YE",
                                  a3_metadata,
                                  "_",
                                  data_stratum_id))

RAW RESULTS

from analysisplan map to outputs ==> analysisplan_outputs
- Recoded data
- analysisplan
- weighting
- cluster id variable name
- questionnaire

results <- from_analysisplan_map_to_output(response, analysisplan = analysisplan[1:20,],
                                           weighting = strata_weight_fun,
                                           cluster_variable_name = NULL,
                                           questionnaire)

FINAL OUTPUTS

individual results
- map_to_visualisation
- map_to_table
- map_to_labeled
list of results (analysisplan_outputs$results)
- lapply(map_to_visualisation)
- lapply(map_to_table)
- lapply(map_to_labeled)
analysisplan_outputs
- hypegrammaR:::map_to_generic_hierarchical_html
  - FUNCTION TO TURN RESULT INTO MARKDOWN
    - hypegrammaR:::from_result_map_to_md_table
    - map_to_md_hypothesis_test(result)
  - analysisplan columns you want to use to structure output
  - prefixes for these when they become headlines
  - level = 2
  - questionnaire
  - output directory
  - output file

map_to_hierarchical_template(results,
                             questionnaire = questionnaire,
                             by_analysisplan_columns =c("dependent.var"
                                                        # ,"repeat.var"
                                                        # ,"repeat.var.value"
                                                        # ,"independent.var",
                                                        # ,"independent.var.value"
                             ),
                             filename = "./output/test.html")


browseURL("./output/test.html")

map_to_master_table(results,"test.csv")