Skip to contents

Make analysis

Source: R/analysis.R
make_analysis.Rd

This function still is experimental. [Experimental]

Usage

make_analysis(
  design,
  survey,
  choices,
  col,
  analysis,
  none_label = NULL,
  group = NULL,
  level = 0.9,
  na_rm = TRUE,
  vartype = "ci",
  get_label = TRUE
)

Arguments

design

A design object

survey

The survey sheet from Kobo that contains at least column 'list_name' (split from 'type') and 'name'

choices

The choices sheet from Kobo contains at least column 'list_name' (split from 'type') and 'name'

col

Column to make analysis from

analysis

One of "median", "mean", "prop_simple", "prop_simple_overall", "prop_multiple", "prop_multiple_overall", "ratio"

none_label

Label for recoding NA if "prop_simple_overall" is selected. If NULL, the code "none_prop_simple_overall" is used as a label.

group

A grouping variable, quoted

level

Confidence level to pass to svy_* functions

na_rm

Should NAs be removed prior to calculation ?

vartype

Parameter from srvyr functions. Default to "ci"

get_label

Should label(s) be joined? Default to TRUE

Value

A summarized analysis

General information

  • Survey and choices must be the final recoded versions of the data. For instance if you have recoded some "other" answers to new choices in the dataset. It must have been added to the choices sheet of the Kobo tool.

  • Design is simply a design object mapped from the dataset thanks to srvyr::as_survey_design().

  • Variables colnames must follow the following pattern in order for

Types of analysis

  • Median: "median" computes the weighted median using svy_median() under the hood

  • Mean : "mean" computes the weighted mean using svy_mean() under the hood

  • Count numeric : "count_numeric" considers a numeric variable as a character one and then computes a simple proportion out of it.

  • Simple proportion : there are two different possible calculation. The first one "prop_simple" removes NA values and calculate the weighted proportion thanks to svy_prop(). The second one "prop_simple_overall" mutate NA values to "none_prop_simple_overall" and then calculates the weighted proportion.

  • Multiple proportion : there are two different possible calculation. The first one "prop_multiple" removes NA values from each dummy 1/0 choice column and calculate the weighted proportion thanks to svy_prop(). The second one "prop_multiple_overall" mutate NA values to 0 for each dummy 1/0 choice column and then calculates the weighted proportion.

  • Ratio: ratio is still under construction for managing NAs. For now it removes them and simply computes the ratio of numeric columns col1 over col2, when col is "col1,col2".