tq_performance cannot pass vector for Rf, PerformanceAnalytics does

[apsteinmetz]

In the performance analytics package, "Rf" is a variable that can be passed as a single value which tq_performance accepts, or a vector, which tq_performance does not.

set.seed(12345)
RaRf<-tibble(date=as.Date(as.yearmon("2009-01-01")+seq(1/12,3,1/12),frac=1),
       Ra=rnorm(3*12,.06/12,.01),
       Rf=rnorm(3*12,.02/12,.001))


RaRf_xts<-as_xts(RaRf,date_col = date)

#in tidyquant -------------------------
# vector for Rf NOT accepted
RaRf %>%
  tq_performance(Ra = Ra, Rf = Rf, performance_fun = SharpeRatio)

#Error in tq_performance_.tbl_df(data = data, Ra = lazyeval::expr_text(Ra),  : 
#object 'ret' not found
#In addition: Warning message:
#  In fun_performance(eval(parse(text = Ra)), ...) : object 'Rf' not found

#in PerformanceAnalytics -------------------------
# vector for Rf accepted
SharpeRatio(RaRf_xts$Ra,RaRF_xts$Rf)
                                       Ra
#StdDev Sharpe (Rf=0.2%, p=95%): 0.5956861
#VaR Sharpe (Rf=0.2%, p=95%):    0.6704678
#ES Sharpe (Rf=0.2%, p=95%):     0.4820222

[mdancho84]

The issue here is the second argument Rf should be Rb for tq_performance. Try this:

RaRf %>%
    tq_performance(Ra = Ra, Rb = Rf, performance_fun = SharpeRatio)

[apsteinmetz]

How does that work for functions that take both Rb and Rf arguments, like AppraisalRatio?

[mdancho84]

It works by selecting the column names as your first two arguments and the performance analytics function as the third.

[apsteinmetz]

But we need 3 column names for AppraisalRatio.

> args(AppraisalRatio)
#function (Ra, Rb, Rf = 0, method = c("appraisal", "modified", 
#    "alternative"), ...) 
NULL

Excuse me if I'm being dense.

[mdancho84]

Normally the Rf is provided as a scalar. Are you suggesting we should also include an option for a column?

[apsteinmetz]

Exactly. Typically, I retrieve the risk-free rate as a series along with everything else. The Performance Analytics package takes it as a scalar OR a series so it's easy to pass them all along. That said, it's no big deal to convert the series to a scalar, like so:

Rf_scalar <-Rf_series %>% tq_performance(Ra=Rf_series, performance_fun = Return.annualized) %>%
  as.numeric()/12  #assuming we want monthly return.

As a practical issue, the fact that we've had very stable short-term rates for years means it doesn't really matter. Indeed, those rates have been about zero so just omitting Rf works too. High and volatile short-term rates might produce different results between a scalar and a vector for some of these stats.