# Use cosinor regression to estimate the phase shift induced by a stimulus during a circadian time-course.

Source:`R/tipa.R`

`tipaCosinor.Rd`

Calculate the phase shift based on fitting sine curves to waveform data
before and after the stimulus, accounting for possible period changes and for
the point in the circadian cycle at which the stimulus occurred. This
function will work best for measurements whose rhythms are approximately
sinusoidal, or at least smoothly increasing and decreasing. If your data are
not sinusoidal, you can first define the phase reference points and then use
`tipaPhaseRef()`

.

## Usage

```
tipaCosinor(
time,
y,
stimOnset,
stimDuration = 0,
periodGuess = 24,
trend = TRUE,
shortcut = TRUE
)
```

## Arguments

- time
Vector of time values for the full time-course.

- y
Vector of measurements (e.g., bioluminescence) for the full time-course.

- stimOnset
Time at which the stimulus started.

- stimDuration
Duration of the stimulus and any transients. Data between

`stimOnset`

and`stimOnset + stimDuration`

will be ignored.- periodGuess
Approximate period of the oscillations (in the same units used in

`time`

), used as initial value in fitting the sine curves.- trend
Model a long-term trend in the cosinor fit for each epoch. Uses a natural cubic spline with 4 degrees of freedom. It is strongly recommended to keep as

`TRUE`

. If set to`FALSE`

, the function may give an error or give completely invalid results.- shortcut
Calculate phase shift using the standard TIPA procedure or using a shortcut based on the phases of the sine curve fits. The two methods give exactly the same result.

## Value

A list.

- phaseShift
Estimated phase shift in circadian hours. Negative values correspond to a delay, positive values an advance.

- epochInfo
Dataframe containing information about the sine curve fits for each epoch: period (in the same units used in

`time`

), phase (in radians), and root mean square error (in the same units as`y`

). If the RMS errors pre-stimulus and post-stimulus are substantially different, then the stimulus may have induced a change in the waveform and thus phase shift estimates may be invalid.

## Examples

```
# Time-course data from multiple (simulated) experiments
getTimecourseFile = function() {
system.file('extdata', 'timecourses.csv', package = 'tipa')}
df = read.csv(getTimecourseFile(), stringsAsFactors = FALSE)
resultList = lapply(sort(unique(df$expId)), function(ii) {
time = df$time[df$expId == ii]
y = df$intensity[df$expId == ii]
tipaCosinor(time, y, stimOnset = 0)})
phaseShifts = sapply(resultList, function(r) r$phaseShift)
```