A set of functions that I have found very useful for speeding up my analysis work. These are well designed (imo) functions that provide a fast alternative to conjuring up functions yourself. These functions are free to use, I don't care (mostly) what you do with it, are provided without any warranty yada yada, and do not give you the right to use it in closed source products (good luck doing that with R code though). Read the LICENSE.
This is a useful case where you are required to aggregate value by a date index into separate periods, possibily for use in tools that work better without pivoted up data (eg. Excel). The syntax is:
agp(table_name, j, by, index)
# j, by have their usual meaning as in data.table
# index is the date column you want to aggregate by
The data can be aggregated for, say, recent 3, 6, 12 months and year to date by suitably defining the values of the date_period list
(check the code for this variable definition). For example, for the following glimpse of data:
> DT
dates grp_1 grp_2 sales
1: 2019-01-01 India Buttons 5.857764
2: 2019-01-08 Brazil Salt 94.159590
3: 2019-01-15 China Buttons 95.128017
4: 2019-01-22 Russia Salt 17.057105
5: 2019-01-29 South Africa Buttons 68.489827
---
99999996: 2022-09-27 India Salt 43.997368
99999997: 2022-10-04 Brazil Buttons 46.981903
99999998: 2022-10-11 China Salt 7.700367
99999999: 2022-10-18 Russia Buttons 84.551482
100000000: 2022-10-25 South Africa Salt 75.611129
the summarized dataset will look something like:
> agp_DT(DT, "sales", c("grp_1", "grp_2"), "dates")
grp_1 grp_2 sales_r3m sales_p3m sales_r6m sales_p6m sales_ytd
1: Brazil Buttons 24986904 50009826 74996730 50008718 262554380
2: Brazil Salt 24996211 24986346 49982557 75021112 245026785
3: China Buttons 25022494 24988808 50011302 74985508 245044854
4: China Salt 49967206 25022877 74990082 50008190 262566617
5: India Buttons 25030072 24992465 50022537 75059273 245038138
6: India Salt 24998593 49995222 74993814 50016927 262498605
7: Russia Buttons 50015343 24993829 75009172 75012270 262531196
8: Russia Salt 24980912 25010533 49991445 74935847 262490950
9: South Africa Buttons 24999193 50004908 75004101 50006625 262463146
10: South Africa Salt 50010186 24989640 74999826 75005350 262494492
Often, in sales analytics, one is required to group certain continous variables into buckets that contribute to 10% of overall sales, in decreasing order of contribution. That is, in the first bucket, you'll have very few members that, together contribute to 10%, and this member count increases as the bucket count increases, each contributing to 10% of the overall sales. In terms of percentiles, the first bucket can be defined as the members that are in the top 90%ile of contributors, the second bucket will be those that are in the top 80%ile of contributors, excluding the top 90%ile of contributors, and so on.
decile_dt(x, ile = 10, decreasing = TRUE)
The decile function defaults to 10 groups, but can be changed to be any n groups i.e. quartlie or quintile through the ile argument.
A quick hack-y replacement for dplyr::glimpse for data.table, but enhanced to differentiate data types by color and read numeric and integer values in human friendly styles (i.e. 4,721,123 as 4.72M). The colours are meant for differentiation, but not identifcation i.e. serves as an analytical aid while reading very long tables. Currently uses a bit too many libraries (4 of them), which I'm not comfortable with - the aim is to reduce them all to 2 (i.e. data.table and crayon, both of which have no dependencies themselves). Stay tuned - this is still very much a WIP.
