elastacloud / automatic-data-explorer Goto Github PK

Functionality to auto write reports with R Markdown. Current idea is to use .Rmd template files that can be chosen by the user and will be automatically filled in depending on values provided by the user to the function arguments.

Change the correlation matrix plot function so that it can plot directly from an input dataframe rather than having to calculate the correlation matrix separately

Correlation functions are not currently being added to the NAMESPACE, add @export to roxygen2 comments

Add packages to DESCRIPTION file so that they should be auto-installed when any user installs the package

Assess unit test coverage and write tests where required

I envisage that this will replace the current method which is an .Rmd template.

This new method would have different scripts, depending on the data type, which are then written to .Rmd, in order, by the autoMarkdown function.

Add option for user to override automatically computed optimal epsilon and minimum points parameters for dbscan

Currently function replaces NAs with the mean of the column. Add option to use na.omit instead.

The current function detects outliers using IQR. Add function that detects outliers using Mahalanobis distance. Add option for user to to specify to use other method than default IQR method.

Add more to the univariate report template

A function that will plot a correlation matrix, with the options of making the plot interactive

Many functions either have no or incomplete help documenation

Some functions currently do not take a dataframe as input. Where possible functions should be able to take a dataframe as input and assess the whole in one call

Add error handling to these functions, more unit-tests and better documentation

Automatically turn a pre existing R script into a .Rmd file that can be rendered into html pdf etc. Needs a way to separate code into 'chunks'. Should be able to reuse the insertChunk and insertQuietChunk functions.

Check for non-numeric data being passed into the function to prevent it from erroring on the cor call

Currently the function will return all correlations between the target variable and the other variables in the function. Add an argument that will allow the user to return only the top N largest correlations.