summary.transformedTrain {STAR} | R Documentation |
Generates a concise summary of transformedTrain
objects. It is mostly
intended for use in batch processing situations where a decision to stop with the
current model or go on with a more complicated one must be made automatically.
## S3 method for class 'transformedTrain': summary(object, ...)
object |
a transformedTrain object. |
... |
additional arguments passed to varianceTime . |
summary.transformedTrain
computes summary statistics
corresponding to plot 1, 2 and 5 of
plot.transformedTrain
.
The first plot tests the uniformity of the spikes (transformed) times
on the (transformed) observation window using a KS test. If the
ecdf of the (transformed) times is within
the 95% band then the first element of component uniformOnTTime
of the returned list is set to TRUE
. It is set to FALSE
otherwise. The second component is relative to the 99% band.
The second plot tests the exponential distribution of the intervals
between successive spikes transformed times. Again if the empirical
curve stays within the 95, respectively 99%, confidence band, the
first, respectively second, element of component BermanTest
of the returned list is set to TRUE
. It is set to FALSE
otherwise.
The fifth plot tests that the variance is equal to the length of the
(transformed) observation time for object
, using point-wise
CI. If n different observation times are defined over the whole
observation window, we expect (1 - CI/100)*n points
to be out with an approximate binomial distribution. For each CI
defined (95 and 99%, by default), component VarTime
of the returned list contains the probability of observing a number as
large as or smaller than the one observed under the binomial null hypothesis.
A list
with the following 3 components:
uniformOnTTime |
A two named components vector of boolean. |
BermanTest |
A two named components vector of boolean. |
VarTime |
A named component vector with as many components as
passed to varianceTime via the ... argument
with p-values of a binomial distribution. |
Christophe Pouzat christophe.pouzat@gmail.com
Ogata, Yosihiko (1988) Statistical Models for Earthquake Occurrences and Residual Analysis for Point Processes. Journal of the American Statistical Association 83: 9-27.
Brown, E. N., Barbieri, R., Ventura, V., Kass, R. E. and Frank, L. M. (2002) The time-rescaling theorem and its application to neural spike train data analysis. Neural Computation 14: 325-346.
transformedTrain
,
plot.transformedTrain
,
mkGLMdf
## Not run: ## Let us consider neuron 1 of the CAL2S data set data(CAL2S) CAL2S <- lapply(CAL2S,as.spikeTrain) CAL2S[["neuron 1"]] renewalTestPlot(CAL2S[["neuron 1"]]) summary(CAL2S[["neuron 1"]]) ## Make a data frame with a 4 ms time resolution cal2Sdf <- mkGLMdf(CAL2S,0.004,0,60) ## keep the part relative to neuron 1, 2 and 3 separately n1.cal2sDF <- cal2Sdf[cal2Sdf$neuron=="1",] n2.cal2sDF <- cal2Sdf[cal2Sdf$neuron=="2",] n3.cal2sDF <- cal2Sdf[cal2Sdf$neuron=="3",] ## remove unnecessary data rm(cal2Sdf) ## Extract the elapsed time since the second to last and ## third to last for neuron 1. Normalise the result. n1.cal2sDF[c("rlN.1","rsN.1","rtN.1")] <- brt4df(n1.cal2sDF,"lN.1",2,c("rlN.1","rsN.1","rtN.1")) ## load mgcv library library(mgcv) ## fit a model with a tensorial product involving the last ## three spikes and using a cubic spline basis for the last two ## To gain time use a fixed df regression spline n1S.fitA <- gam(event ~ te(rlN.1,rsN.1,bs="cr",fx=TRUE) + rtN.1,data=n1.cal2sDF,family=binomial(link="logit")) ## transform time N1.Lambda <- transformedTrain(n1S.fitA) ## check out the resulting spike train using the fact ## that transformedTrain objects inherit from spikeTrain ## objects N1.Lambda ## Use more formal checks summary(N1.Lambda) plot(N1.Lambda,which=c(1,2,4,5),ask=FALSE) ## Transform spike trains of neuron 2 and 3 N2.Lambda <- transformedTrain(n1S.fitA,n2.cal2sDF$event) N3.Lambda <- transformedTrain(n1S.fitA,n3.cal2sDF$event) ## Check interactions summary(N2.Lambda %frt% N1.Lambda) summary(N3.Lambda %frt% N1.Lambda) plot(N2.Lambda %frt% N1.Lambda,ask=FALSE) plot(N3.Lambda %frt% N1.Lambda,ask=FALSE) ## End(Not run)