tidyquant with grouped data frame? about tidyquant

[mdancho84]

This is definitely a bug with grouped data frames. The good news is it can be fixed. Look for a patch soon.

[mdancho84]

@DavisVaughan

Glad it's working. I will close out. Regarding your comments, let me see what I can do. Here's my specific comments:

1. I will incorporate in 0.3.0. The tough part about column names is that sometimes there is more than one column being added, which is why mutate doesn't work with say BBands. I could add a generic column name field, and as long as the user enters the appropriate number of column names using a character vector, col_names = c("col_one", "col_two"), then this would work. However, if the correct number of column names is not specified I guess the user gets the default generated comments.

2. Yeah, this is the problem with inconsistent functions, especially with the na.pad of fill arguments. I'll see what I can do, but it may be best to leave it up to the user to get it right in this instance.

[DavisVaughan]

Fixing code block output. Didn't render correctly first time. My bad.

[mdancho84]

Davis,

Good pickup. Let me check this out tomorrow AM.

Thanks, Matt

[mdancho84]

@DavisVaughan

I've updated the development version with a patch that enables tq_mutate and tq_transform to work seamlessly with dplyr::group_by. The tests I've run are as follows:

• Tested using group_by with both tq_mutate and tq_transform to ensure results are the same as using group_by with mutate from dplyr (note that tq_transform will only return the group column(s), date column and new column(s); tq_mutate returns the same result as mutate)
• Tested using your example
• Tested with zero, one, and multiple groups

Please test it out. Make sure it works to your satisfaction. I aim to please, so let me know if there is still something that needs worked out. If it's all good, let me know and I will close out the issue.

You can download the development version with devtools:

# install.packages("devtools") # uncomment if not installed
devtools::install_github("mdancho84/tidyquant")

Also, you may need to update some of the dependencies prior to installing the dev version of tidyquant. I had to update tidyr before I ran the devtools::install_github("mdancho84/tidyquant"). If using RStudio, just "update all" packages prior to updating just to be on the safe side. Otherwise, you may get an error when downloading due to a conflict.

[DavisVaughan]

@mdancho84

This is great! That example works like a charm now. In terms of this specific issue, you can close it.

I did have two other ideas though, just something that I thought I might bring to your attention.

• When running the code y %>% tq_mutate(ohlc_fun = Ad, mutate_fun = runSD), the resulting tibble has v1 as the new column name. y %>% mutate(sd = runSD(adjusted)) allows me to specify sd as the column name, which would be efficient if more calculations were to be immediately done on the sd column through piping. Obviously the same sd = technique is not going to work in your case, but perhaps another parameter such as col_name = or var_name would work to where the user could specify the name of the resulting column in the tq_mutate call. Just a thought and really a minor quirk.

• Some functions like rollmean don't work right out of the box if you are attempting to use them with tq_mutate. Specifically, using the grouped y from the example above, and running y %>% tq_mutate(ohlc_fun = Ad, mutate_fun = rollmean, k = 10) returns an error that two tibbles can't be joined. rollmean is doing what it is supposed to do by default, by not padding with NAs so the result is shorter in length than the original tibble. However, some people might not be familiar with that, and get frustrated, unaware that y %>% tq_mutate(ohlc_fun = Ad, mutate_fun = rollmean, k = 10, fill = NA, align = "right") would work. Since tq_mutate is almost always going to be working on a tibble, and the alignment and fill would be needed, perhaps some sensible defaults could be set for fill and align? Again, it's nothing code breaking, but just a personal preference while messing around with the package.

Great work though! And thanks again.