Flexible Seasonality Graphs — fg_seasonalstudy • FinanceGraphs

Flexible and general seasonality graphs.

Usage

fg_seasonalstudy(
  indta,
  seasonaltype = "yr",
  seasonaldateset = NULL,
  day_offset = 0,
  bdaysonly = TRUE,
  graphtype = "flex",
  normalize = "",
  projectfwd = "",
  projfwd_wt = 0.9,
  yvar = "value",
  title = NULL,
  yrange = NULL,
  n_color_switch = 7,
  n_hex_switch = 20,
  line_aes_set = "lines",
  line_on_lastdate = TRUE,
  killbad_eop = FALSE,
  return_dates = FALSE
)

Arguments

indta

A data.frame with at least one date column and a numeric column with the name given by yvar

seasonaltype

(default: "yr") A string denoting the periodicity of the analysis, must be in one of c("yr","qtr","mo","wk",IMMroll","optmo","optqtr"). Not used if seasonaldataset is specified instead

seasonaldateset

(default:NULL) An optional dataset with two columns: A Date column defining each period, and possibly a character variable with a period identifier.

day_offset

(default: 0) Number of days (determined by bdaysonly) to offset each period. Applies only to c("yr","qtr","mo","wk")

bdaysonly

(default:TRUE) Only consider NYSE business days.

graphtype

(default: "flex") One of the following

`graphtype`	Description
`line`	A line for every period, with identifiers placed near the end
`hex`	A density plot for each day of the seasonal period, with the last seasonal period kept as a line
`flex`	A line graph if there are less than `n_hex_switch` seasonal periodss
`stat`	A line graph showing 10tyh, 50th, and 90th percentiles of values for each day in a seasonal period

normalize

(default ""). How to normalize each period. Default is no normalization. Other options are

`normalize`	Description
`relative`	Older seasonal periods adjusted to match beginning of latest period
`index`	Each seasonal period is expressed as index from beginning of each period

projectfwd

(default "") Project forward based on same number of days in seaonal period. Default is no projection.

`normalize`	Description
`mean`	Use mean (by days in seasonal period) of each cumulative percentage change from the start of the period
`weighted`	Weight previous period cumulative percantge change with an expoential decay using `projfwd_wt` ** (Periods back)

projfwd_wt

See above

yvar

(default:"value") Series to use in indta

title

(default: NULL) Title for graph.

yrange

(default: NULL) y axis Range for which to focus data.

n_color_switch

(default 7) Number of periods past which lines will be colored by a descending scale.

n_hex_switch

(default 20) Number of periods past which a hex graph will be used, if flex is chosen above.

line_aes_set

(default "lines") Aes color set for discrete lines.

line_on_lastdate

(default TRUE) Add a vertical line at the last observations day in period.

killbad_eop

(default FALSE): DO not show periods for which there are at least 60% of the mean number of observations per day of the seasonal period. If used, this helps to curtail extreme moves at the end of a period. For example, this would redact the 366th day of the year.

return_dates

(default: FALSE) Return list(graph,dates) instead of just the graph.

Value

a ggplot() object displaying seasonality

Details

Seasonality Graphs

Examples

require(data.table)
assetcols <- c("EEM","IBM","QQQ","TLT")
eqtyidx<-eqtyrtn[,(assetcols):=lapply(.SD,\(x) 100*(exp(cumsum(fcoalesce(x,0))))),
            .SDcols=assetcols]
fg_seasonalstudy(eqtyidx,yvar="TLT",seasonaltype="yr",normalize="index",projectfwd="mean")

# Earnings seasons
earnings_dates <- earnings_ibm[,.(reportedDate,divdt=format(reportedDate,"%Y%m%d"))]
fg_seasonalstudy(eqtyidx,yvar="IBM",seasonaldateset = earnings_dates,graphtype="stat",
   normalize="index",projectfwd="mean")