# NAME

Statistics::Sequences::Turns - Kendall's turning-points test - of peaks and troughs in a numerical sequence

# VERSION

This is documentation for Version 0.13 of Statistics::Sequences::Turns.

# SYNOPSIS

`````` use strict;
use Statistics::Sequences::Turns 0.13;
my \$turns = Statistics::Sequences::Turns->new();
\$turns->load([2, 0, 8.5, 5, 3, 5.01, 2, 2, 3]); # numbers; or give as "data => \$aref" with each stat call
my \$val = \$turns->observed(); # or descriptive methods: expected(), variance(), obsdev() and stdev()
\$val = \$turns->z_value(); # # or in list context get both z- and p-value
\$val = \$turns->p_value(); # as above, assume data are loaded
my \$href = \$turns->stats_hash(values => [qw/observed p_value/], ccorr => 1); # incl. any other stat-method
\$turns->dump(values => [qw/observed expected p_value/], ccorr => 1, flag => 1, precision_s => 3, precision_p => 7);
# prints: observed = 11.000, expected = 10.900, p_value = 0.5700167``````

# DESCRIPTION

Implements Kendall's (1973) "turning point test" of sudden changes as peaks and troughs in the values of a numerical sequence. It is sometimes described as a test of "cyclicity", and often used as a test of randomness. Kendall (1973) introduced this as a test of ups and downs relative to linear progressions in a sequence (ahead of describing tests based on autocorrelation and Fourier analysis).

Specifically, for a sequence of numerical data (interval or ordinal) of size N, a count of turns is incremented if the value on trial i, for all i greater than zero and less than N, is, with respect to its immediate neighbours (the values on i - 1 and i + 1), greater than both neighbours (a peak) or less than both neighbours (a trough). The difference of this observed number from the mean expected number of turns for a randomly generated sequence, taken as a unit of the standard deviation, gives a Z-score for assessing the "randomness" of the sequence, i.e., the absence of a factor systematically affecting the frequency of peaks/troughs, given that, for turns, there is "a fairly rapid tendency of the distribution to normality" (Kendall 1973, p. 24).

With these local fluctuations tested regardless of their spacing and magnitude, the test does not indicate if the changes actually cycle between highs and lows, if they are more or less balanced in magnitude, or if any cycling is periodic; only if oscillation in general is more common than linear progression.

# METHODS

## new

`` \$turns = Statistics::Sequences::Turns->new();``

Returns a new Turns object. Expects/accepts no arguments but the classname.

`````` \$turns->load(@data);
\$turns->load('foodat' => \@data); # labelled whatever``````

Loads data anonymously or by name - see load in the Statistics::Data manpage for details on the various ways data can be loaded and then retrieved (more than shown here). Data must be numerical (ordinal, interval type). All elements must be numerical of the method croaks.

See Statistics::Data for these additional operations on data that have been loaded.

## observed

`````` \$v = \$turns->observed(); # use anonymously loaded data
\$v = \$turns->observed(name => 'mysequence'); # ... or by "name" given on loading
\$v = \$turns->observed(data => \@data); # ... or just give the data now``````

Returns observed number of turns. This is the number of peaks and troughs, starting the count from index 1 of the sequence (a flat array), checking if both its immediate left/right (or past/future) neighbours are lesser than it (a peak) or greater than it (a trough). Wherever the values in successive indices in the sequence are equal, they are treated as a single observation/datum - so the following:

`` 0 0 1 1 0 1 1 1 0 1``

is counted up for turns as

`````` 0 1 0 1 0 1
* * * *``````

This shows four turns - two peaks (0 1 0) and two troughs (1 0 1).

Returns 0 if the given list of is empty, or the number of its elements is less than 3.

## expected

`````` \$v = \$turns->expected(); # use loaded data; or specify by "name"
\$v = \$turns->expected(data => \@data); # use these data
\$v = \$turns->expected(trials => POS_INT); # don't use actual data; calculate from this number of trials``````

Returns the expected number of turns, which is set by N the number of trials/observations/sample-size ...:

E[T] = 2 / 3 (N – 2)

or, equivalently (in some sources),

E[T] = ( 2N – 4 ) / 3

## variance

`````` \$v = \$turns->variance(); # use loaded data; or specify by "name"
\$v = \$turns->variance(data => \@data); # use these data
\$v = \$turns->variance(trials => POS_INT); # don't use actual data; calculate from this number of trials``````

Returns the expected variance in the number of turns for the given length of data N.

V[T] = (16N – 29 ) / 90

## obsdev

`````` \$v = \$turns->obsdev(); # use data already loaded - anonymously; or specify its "name"
\$v = \$turns->obsdev(data => \@data); # use these data``````

Returns the observed deviation from expectation for the loaded/given sequence: observed less expected turn-count (O - E). Alias of `observed_deviation` is supported.

## stdev

`````` \$v = \$turns->stdev(); # use data already loaded - anonymously; or specify its "name"
\$v = \$turns->stdev(data => \@data);``````

Returns square-root of the variance. Aliases `standard_deviation` and `stddev` (common in other Statistics modules) are supported.

## z_value

`````` \$z = \$turns->z_value(ccorr => 1); # use data already loaded - anonymously; or specify its "name"
\$z = \$turns->z_value(data => \$aref, ccorr => BOOL);
(\$z, \$p) = \$turns->z_value(data => \$aref, ccorr => BOOL, tails => 2); # same but wanting an array, get the p-value too``````

Returns the deviation ratio, or Z-score, taking the turncount expected from that observed and dividing by the root variance, by default with a continuity correction in the numerator. Called in list context, returns the Z-score with its normal distribution, two-tailed p-value.

The data to test can already have been loaded, or sent directly as an aref keyed as data.

Optional named arguments tails (1 or 2), ccorr (Boolean for the continuity-correction), precision_s (for the statistic, i.e., Z-score) and precision_p (for the p-value).

The method can all be called with "sufficient" data: giving, instead of actual data, the observed number of turns, and the number of trials, the latter being sufficient to compute the expected number of turns and its variance.

Alias `z_score` is supported.

## p_value

`````` \$p = \$turns->p_value(); # using loaded data and default args
\$p = \$turns->p_value(ccorr => BOOL, tails => 1|2); # normal-approximation based on loaded data
\$p = \$turns->p_value(data => \$aref, ccorr => BOOL, tails => 2); #  using given data``````

Returns the normal distribution p-value for the deviation ratio (Z-score) of the observed number of turns, 2-tailed and continuity-correct by default (or set tails => 1 and ccorr => 0, respectively). Other arguments are as for z_value.

## stats_hash

`` \$href = \$turns->stats_hash(values => [qw/observed expected variance z_value p_value/], ccorr => 1);``

Returns a hashref for the counts and stats as specified in its "values" argument, and with any options for calculating them. See stats_hash in the Statistics::Sequences manpage for details/options. If calling via a "turns" object, the option "stat => 'turns'" is not needed (unlike when using the parent Statistics::Sequences object).

## dump

`` \$turns->dump(flag => BOOL, verbose => BOOL, format =>  'table|labline|csv');``

Print test results to STDOUT. See dump in the Statistics::Sequences manpage for details/options.

# EXAMPLE

## Seating at the diner

This is the data from Swed and Eisenhart (1943) also given as an example for the Runs test, Joins test and Vnomes (serial) test. It lists the occupied (O) and empty (E) seats in a row at a lunch counter. Have people taken up their seats on a random basis - or do they show some social phobia (more sparsely seated than "chance"), or are they trying to pick up (more compactly seated than "chance")? What does Kendall's test of turns reveal?

`````` use Statistics::Sequences::Turns;
my \$turns = Statistics::Sequences::Turns->new();
# change the nominal data from Swed & Eisenhart (1943) into numerical values:
my @seating = map { \$_ eq 'E' ? 1 : 0 } (qw/E O E E O E E E O E E E O E O E/);
\$turns->dump_vals(delim => q{,}); # via Statistics::Data - prints the 1s and 0s:
# 1,0,1,1,0,1,1,1,0,1,1,1,0,1,0,1
\$turns->dump(
format => 'labline',
flag => 1,
precision_s => 3,
precision_p => 3,
verbose => 1,
);``````

This prints:

`` Turns: observed = 9.000, p_value = 0.050``

So, the observed number of turns in the seating arrangements differed from that expected within the bounds of chance, at the .05 level. The Vnomes test for trinomes was similarly marginal (p = .044), as was the result for Runs (p = 0.055), while the Joins test was clearly non-significant (p = .302). Checking the number of turns expected ( = 6) suggests, perhaps, a tendency for people to take their seats further away from each other (leave more unoccupied seats between them) than expected on the basis of chance.

# DEPENDENCIES

Statistics::Sequences

Statistics::Zed

Kendall, M. G. (1973). Time-series. London, UK: Griffin. ISBN 0852642202. [The test is described on pages 22-24 of the 1973 edition; in the Example 2.1 for this test, the expected number of turns should be calculated with the value 52 (i.e., with N - 2), not the misprinted value of 54.]

Statistics::Sequences for other tests of sequences, and for sharing data between these tests.

Statistics::Sequences::Joins : another test of consecutive values in a sequence, examining alternations.

Statistics::Sequences::Pot : another trend-type test, examining relatively spaced clustering of particular events.

# SUPPORT

You can find documentation for this module with the perldoc command.

``    perldoc Statistics::Sequences::Turns``

You can also look for information at: