Biology Reference
In-Depth Information
The file WT.NO represents 14 days of free-running locomotor activity of
the wild-type hamster. The data were accumulated from an animal, under
experimental conditions of DD, following a long-term regimen (2-week
minimum) of 12 hours of light and 12 hours of dark (an LD 12:12 cycle).
The file TAUSS.NO represents 14 days of free-running locomotor
activity of a tau
ss
hamster. The data have been accumulated from an
animal, under experimental conditions of DD, following a long-term
exposure (two-week minimum) to 10 hours of light and 10 hours of dark
(an LD 10:10 cycle).
The file TRNSPLNT.NO represents 14 days of free-running locomotor
activity of the SCN transplant recipient.
We begin by plotting the data. The plots themselves make apparent some
of the challenges in analyzing this simulated data. Clearly, all the plots
exhibit a rhythmic pattern, with the pattern in the first data set being,
perhaps, best expressed. Notice, however, that the rhythms are
confounded. For the first two plots, the amplitudes change with time, and
a shift that also grows with time is visible in the position of the peaks of the
repeating patterns. In addition, the presence of noise makes it difficult to
visually identify the exact location of the peaks. For the third plot, a
rhythmic pattern is also clearly present, but it is more subtle. In order to
address the questions raised above, we shall need to quantify the rhythms
as accurately as possible and then analyze them appropriately.
This example illustrates some of the general analytical challenges
presented by confounded time series, namely: (1) Mean and/or variance
nonstationarities (i.e., time-dependent drifting and/or changes in
oscillatory amplitude); (2) period and/or phase instability; and (3) noise.
To overcome these challenges, the following question is important:
What do we wish to learn from our data to be able to quantitatively
characterize it? The typical analytical objectives of rhythmic analyses are
to extract information about: (1) The period of expression; (2) the phase
of expression; (3) the oscillatory amplitude of expression; and (4) the
robustness of rhythmic expression (how strongly rhythmic the observed
patterning is).
Even from the simple examples in Figure 11-9, it is clear that to answer
these questions and the specific questions raised in the case study, we
shall need tools that will allow us to work with confounded data. We
would like (1) detrending strategies (to address mean nonstationarities,
or drifting data); (2) strategies for data normalization (to address
variance nonstationarities, or variable-dynamic-range data); and
(3) analytical algorithms that, by design, attempt to accommodate the
nonstationarities that may be present in uncorrected data series. Before
proceeding with the example, we describe some tools for analyzing
confounded time series. We begin by outlining some well-known
fundamentals.
