calcSDA: Calculate Specific Dynamic Action for a fish or aquatic...
In denis-chabot/fishMO2: Calculate and Plot SMR, O2crit and SDA, and convert between oxygen units

Description Usage Arguments Details Value Warning Author(s) References See Also Examples

Calculate the peak value, the time-to-peak, the duration and the magnitude of the Specific Dynamic Action (SDA, i.e., Heat Increment of Feeding, post-prandial metabolic rate, etc.) from oxygen uptake data recorded with a fed fish for which the standard metabolic rate (SMR) is known. SMR is required to calculate the magnitude of SDA (area bounded by the SDA curve and the SMR). A non-parametric quantile regression is used to fit the SDA curve, a technique that is robust against changes in activity level during digestion.

calcSDA(
  X,
  Y,
  my.smr,
  tau = 0.2,
  lambda = 24,
  postfeed.acclim = 6,
  tol = 0.05,
  tol.type = c("proportion", "constant"),
  MO2.time.unit = c("hour", "min", "day"),
  X.time.unit = c("hour", "min", "day")
)

`X`	A numeric vector or data frame column containing the time relative to the meal (which it at time zero) for each oxygen uptake measurement. Measurements obtained before the meal, if present, have negative X values. See `X.time.unit`.
`Y`	A numeric vector or data frame column containing oxygen uptake measurements.
`my.smr`	The Standard Metabolic Rate (SMR) for the animal. Same units as `Y`.
`tau`	A parameter that varies between 0 and 1. As with a quantile, the `rqss` function fits a curve that results in the probability of the response, Y, of falling below the line is tau, and of falling above is 1 − tau. If SMR has been calculated using a quantile, it is logical to make tau == p, and often p is 0.2.
`lambda`	A penalty parameter used by `rqss` to control the flexibility of the fitted curve. Assuming that X is in hours, this should be a number of hours greater than the period of activity cycles. Typically such cycles are circadian and values of lambda between 18 and 30 work well.
`postfeed.acclim`	Duration (in same units as X axis, typically hours) of a period, right after feeding, which is skipped before fitting the SDA curve with `rqss`. See `Details`.
`tol`	How close to SMR should the SDA curve return to declare SDA complete. Ideally the curve returns to SMR (tol == 0). In practice, small fluctuations in MO2 around SMR, or experiments that end a bit too early for the fit to return to SMR will result in curve that do not return to SMR and SDA will have no end, no duration, and the magnitude will not be reliable. A solution is to allow for a tolerance about SMR to declare that SDA has ended. A good value is 5% of SMR, expressed as a proportion (0.05).
`tol.type`	If `proportion`, the value set by `tol` is interpreted as a percent of SMR. If set to `constant`, a set value (same scale as MO2) is added to SMR.
`MO2.time.unit`	Time unit over which MO2 is expressed, `hour, min, day`. Required to integrate the area under the curve bound by SDA and SMR.
`X.time.unit`	Time can be in `hour, min, day` units, but will be transformed into hours within this function.

The rqss function of the quantreg package is used for this non-parametric quantile regression that results in a curve that tracks the bottom of the MO2 values on a plot of MO2 x time, keeping a proportion tau of values below the line. Thus the same assumption is made than when estimating SMR: SMR is not taken as the lowest value of MO2 that can be observed, but as a normal distribution describing the MO2 when an animal is fasted and quiet, and some values are below the mean of this distribution. In practice, when measuring SMR, a method that fits several normal distributions, should reveal SMR, which is the lowest of these distributions. In practice, it does not work in all cases (See Chabot et al. 2016b) and simply assuming that 20% of the MO2 values are below SMR works well. This same principle is used to fit a curve tracing SDA. Without ignoring a few hours at the beginning of SDA, it is difficult to get a fit that begins close to SMR at time zero. The practical, if not particularly elegant, solution is to start the fit a few hours hours after the meal and to force a straight line from SMR at time zero, to the beginning of the SDA fit at time postfeed.acclim.

A list with two object:

sda.fit

A data frame containing the fitted values of SDA every 0.25 hour, which can then be used to plot SDA. The variables of this data frame are:

time: A numeric in hours, spaced every 0.25 h except between time 0 and time postfeed.acclim.
pred: Predicted MO2 (including SMR).
net: Predicted MO2 (after removing SMR).
status: Status of each predicted MO2 value, either start, sda, post-sda.

sda.var

A list that contains the calculated parameters of SDA:

duration: Duration of SDA in hours.
peak.time: Time in hours when the peak of SDA is reached, with 0 == time of feeding.
peak: Maximum MO2 due to SDA (including SMR). Same unit as Y.
peak.net: Maximum MO2 due to SDA (after removing SMR). Same unit as Y.
magnitude: Area under the curve representing SDA and above the straight horizontal line representing SMR. Same units as Y without the time component.
end.MO2: If the fit reached SMR + tol, SDA has ended and the value of tol is entered, expressed as a proportion of SMR. If the SDA has not ended properly, the difference between the lowest observed value of SDA and SMR is shown, expressed as a proportion of SMR.
SMR: The value of SMR used to calculate SDA.
tau: The value of tau used to calculate SDA.
lambda: The value of lambda used to calculate SDA.

The parameters tau and lambda are not independent from each other. If changing one, check if the results are acceptable, otherwise adjust the second parameter.

Denis Chabot, Institut Maurice-Lamontagne, Department of Fisheries and Oceans.

Chabot, Denis, Koenker, Roger and Farrell, Anthony P. (2016a) The measurement of the specific dynamic action in fishes. Journal of Fish Biology 88, 152-172.

Chabot, Denis, Steffensen, John F. and Farrell, Anthony P. (2016b) The determination of the standard metabolic rate in fishes. Journal of Fish Biology 88, 81-121.

Koenker, Roger (2005) Quantile regression. Cambridge University Press, Cambridge. 366 p.

calcSMR, rqss

# using juvenile cod data included in the package, minimum acceptable r2 is
# 0.96 and SMR is 81.2; see calcSMR for how it was obtained
data("codSDA")
sda.data = subset(codSDA, r2 >= 0.96)
# Logtime_hr is time in hours relative to time of feeding, a required variable.
# MO2cor is in micromoles per min per kg and the "cor" in the name means that
# the MO2 values soon after feeding were corrected for the effect of the feeding procedure

SDA= calcSDA(X=sda.data$Logtime_hr, Y=sda.data$MO2cor, my.smr=81.2,
               tau=0.2, lambda=30,
	           postfeed.acclim=3, tol=0.05, tol.type="proportion",
	           MO2.time.unit="min", X.time.unit="hour")
names(SDA)
SDA$sda.var