Von Neuman et al. (1941) discussed the Mean Square of Successive Differences as a measure of variability that takes into account gradual shifts in mean. This is appropriate when studying errors in ballistics or variability and stability in mood when studying affect. For random data, this will be twice the variance, but for data with a sequential order and a positive autocorrelation, this will be much smaller. Since the mssd is just twice the variance - the autocorrelation, it is thus possible to also find the autocorrelation for a particular lag.
mssd(x,group=NULL, lag = 1,na.rm=TRUE)
rmssd(x,group=NULL, lag=1, na.rm=TRUE)
autoR(x,group=NULL,lag=1,na.rm=TRUE,use="pairwise")The average squared successive difference (mssd) and the square root of the average squared successive difference (rmssd). Note that this is not the same as the standard deviation of the lagged differences.
a vector, data.frame or matrix
the lag to use when finding diff
A column of the x data.frame to be used for grouping
Should missing data be removed?
How to handle missing data in autoR
.
William Revelle
When examining multiple measures within subjects, it is sometimes useful to consider the variability of trial by trial observations in addition to the over all between trial variation. The Mean Square of Successive Differences (mssd) and root mean square of successive differences (rmssd) is just
\(\sigma^2 = \Sigma(x_i - x_{i+1})^2 /(n-lag) \)
Where n-lag is used because there are only n-lag cases.
In the case of multiple subjects (groups) with multiple observations per subject (group), specify the grouping variable will produce output for each group.
Similar functions are available in the matrixStats package. However, the varDiff function in that package is variance of the difference rather than the MeanSquare. This is just the variance and standard deviation applied to the result from the diff function.
Perhaps useful when studying mood, the autoR function finds the autocorrelation for each item for the specified lag. It also returns the rmssd (root means square successive difference). This is done by finding the correlation of the lag data.
Jahng, Seungmin and Wood, Phillip K and Trull, Timothy J. Analysis of affective instability in ecological momentary assessment: Indices using successive difference and group comparison via multilevel modeling. Psychological methods (2008) 13, 354-375.
Von Neumann, J., Kent, R., Bellinson, H., and Hart, B. (1941). The mean square successive difference. The Annals of Mathematical Statistics, pages 153-162.
See Also rmssd for the standard deviation or describe for more conventional statistics. describeBy and statsBy give group level statistics. See also link{mlr}, link{mlreliability}, link{mlPlot} for other ways of examining within person variability over time.
t <- seq(-pi, pi, .1)
trial <- 1: length(t)
gr <- trial %% 8
c <- cos(t)
ts <- sample(t,length(t))
cs <- cos(ts)
x.df <- data.frame(trial,gr,t,c,ts,cs)
rmssd(x.df)
rmssd(x.df,gr)
autoR(x.df,gr)
describe(x.df)
#pairs.panels(x.df)
#mlPlot(x.df,grp="gr",Time="t",items=c(4:6))
Run the code above in your browser using DataLab