Skip to content

Groups

Like the cross-sectional operators, but within groups: each stock is compared only with the stocks sharing its label that day. In the toy data, A and B are Tech, C and D are Energy.

groups is a group's name ('sector') or any expression giving labels: quantile(field('mcap'), 5), coalesce(group('sector'), 'Other'), group_cross(...). A stock with no label, or outside the universe, gets no value. Labels given on fewer dates than the prices hold from their date until the next, never earlier. Groups and Labels works through them on an example.

Labels

group

group('name')

The labels themselves. Labels can be compared with == and !=, used as groups, and passed through where and coalesce; anything else (arithmetic, rank) is refused when the expression is read.

group('sector')
ABCD
Jan 02TechTechEnergyEnergy
Jan 03TechTechEnergyEnergy
Jan 04TechTechEnergyEnergy
Jan 05TechTechEnergyEnergy
Jan 08TechTechEnergyEnergy
Jan 09TechTechEnergyEnergy
group('sector') == 'Tech'
ABCD
Jan 021100
Jan 031100
Jan 041100
Jan 051100
Jan 081100
Jan 091100
where(group('sector') == 'Financials', None, field('book_to_price'))    # not for banks

identifier

identifier()

Each stock's own identifier (its name in the data), as a label, on every date. On its own it says nothing new; its use is reading a stock's identity through an operator that moves values between stocks. ts_match(identifier(), key, 252) is each stock's twin's identifier: the way to see the twins (Statistical Twins).

identifier()
AAABBBCCCDDDEEEFFF
Apr 09AAABBBCCCDDDEEEFFF
Apr 10AAABBBCCCDDDEEEFFF
Apr 11AAABBBCCCDDDEEEFFF

group_cross

group_cross(a, b, ...)

One grouping from several: each stock's labels joined with a bar, such as 'Tech|2'. Below, sector crossed with beta halves: the high-beta stock in each sector is in half 2.

group_cross('sector', quantile(field('beta'), 2))
ABCD
Jan 02Tech|2Tech|1Energy|2Energy|1
Jan 03Tech|2Tech|1Energy|2Energy|1
Jan 04Tech|2Tech|1Energy|2Energy|1
Jan 05Tech|2Tech|1Energy|2Energy|1
Jan 08Tech|2Tech|1Energy|2Energy|1
Jan 09Tech|2Tech|1Energy|2Energy|1
group_demean(returns(20), group_cross('sector', quantile(field('mcap'), 3)))    # sector- and size-neutral

Within each group

group_demean

group_demean(x, groups)

x minus its group's mean that day: neutral to the group.

returns(1)
ABCD
Jan 02····
Jan 030.100−0.0500.0170.050
Jan 040.091−0.053·−0.024
Jan 05−0.0830.0280.0160.049
Jan 080.182−0.0810.0320.023
Jan 090.077−0.059−0.0160.045
group_demean(returns(1), 'sector')
ABCD
Jan 02····
Jan 030.075−0.075−0.0170.017
Jan 040.072−0.072·0.000
Jan 05−0.0560.056−0.0160.016
Jan 080.131−0.1310.005−0.005
Jan 090.068−0.068−0.0310.031
group_demean(returns(20), 'sector')                         # sector-neutral momentum
group_demean(returns(20), quantile(field('mcap'), 5))       # size-neutral momentum

group_rank

group_rank(x, groups)

The percentile rank within each group, from 1/n to 1.

returns(1)
ABCD
Jan 02····
Jan 030.100−0.0500.0170.050
Jan 040.091−0.053·−0.024
Jan 05−0.0830.0280.0160.049
Jan 080.182−0.0810.0320.023
Jan 090.077−0.059−0.0160.045
group_rank(returns(1), 'sector')
ABCD
Jan 02····
Jan 031.00.50.51.0
Jan 041.00.5·1.0
Jan 050.51.00.51.0
Jan 081.00.51.00.5
Jan 091.00.50.51.0

group_gaussrank

group_gaussrank(x, groups)

gaussrank within each group: \(\Phi^{-1}((r_i - 0.5)/n)\) with \(r_i\) the stock's rank and \(n\) the number of stocks with a value in its group that day. Normal scores within each sector, summing to zero in each; a stock alone in its group gets 0.

returns(1)
ABCD
Jan 02····
Jan 030.100−0.0500.0170.050
Jan 040.091−0.053·−0.024
Jan 05−0.0830.0280.0160.049
Jan 080.182−0.0810.0320.023
Jan 090.077−0.059−0.0160.045
group_gaussrank(returns(1), 'sector')
ABCD
Jan 02····
Jan 030.674−0.674−0.6740.674
Jan 040.674−0.674·0.000
Jan 05−0.6740.674−0.6740.674
Jan 080.674−0.6740.674−0.674
Jan 090.674−0.674−0.6740.674
group_gaussrank(field('book_to_price'), 'sector') + group_gaussrank(returns(231), 'sector')   # sector-neutral, on one scale

group_quantile

group_quantile(x, groups, n)

The bucket 1 to n within each group: a dependent double sort.

group_quantile(field('value'), quantile(field('mcap'), 3), 3)    # value terciles within size terciles

Group statistics

Every member of a group gets the group's number, even a member whose own value is missing: C has no dollar volume on Jan 04 and still gets the Energy sum.

group_std

group_std(x, groups)

The standard deviation across the group's stocks that day; no value for a group with only one value, as Energy on Jan 04.

returns(1)
ABCD
Jan 02····
Jan 030.100−0.0500.0170.050
Jan 040.091−0.053·−0.024
Jan 05−0.0830.0280.0160.049
Jan 080.182−0.0810.0320.023
Jan 090.077−0.059−0.0160.045
group_std(returns(1), 'sector')
ABCD
Jan 02····
Jan 030.1060.1060.0240.024
Jan 040.1010.101··
Jan 050.0790.0790.0230.023
Jan 080.1860.1860.0060.006
Jan 090.0960.0960.0430.043
group_demean(returns(5), 'sector') / group_std(returns(5), 'sector')    # a z-score within each sector

group_sum

group_sum(x, groups)

The sum within the group; no value when the group has no values that day.

adv(1)
ABCD
Jan 0210006000600016000
Jan 0313205320640515960
Jan 0418004680·20500
Jan 0512105550806018060
Jan 0826004250768019360
Jan 0925203840724527600
group_sum(adv(1), 'sector')
ABCD
Jan 02700070002200022000
Jan 03664066402236522365
Jan 04648064802050020500
Jan 05676067602612026120
Jan 08685068502704027040
Jan 09636063603484534845
adv(20) / group_sum(adv(20), 'sector')         # each stock's share of its sector's trading

group_count_valid

group_count_valid(x, groups, period)

How many values the group's stocks have had over the last period days.

group_count_valid(field('surprise'), 'sector', 63)     # how many reports the sector had this quarter