compPowerRespDetection: Compute power to detect differences from background where some categories differ from background.

Description

Compute power to detect responses differing from background in a simulation of responses to a number of categories of stimuli, some of which differ from a background level of firing.

Usage

compPowerRespDetection(bkgLevel, respLevel, numCats, numCatsWithResp, normDistribution =
FALSE, showProgress = FALSE, numTrialsPerCat = 15, numBootIters = 1000, numRuns = 1000,
alpha = 0.05)

Arguments

bkgLevel

Background firing rate, prior to stimulus onset.

respLevel

Response firing rate during categories which have a response other than background.

numCats

Total number of categories of stimulus.

numCatsWithResp

Number of categories to be given respLevel responses.

normDistribution

TRUE if normally distributed response and background counts should be used, otherwise Poisson distributed counts are used. Default is FALSE.

showProgress

TRUE if run count should be printed. Default is FALSE.

numTrialsPerCat

Number of trials of responses and backgrounds for each category. Default is 15.

numBootIters

Number of boot iterations, passed as 'R' argument to boot function. Default is 1000.

numRuns

Number of simulations to perform followed by bootstrap testing. Default is 1000.

alpha

Significance level, alpha, to use in counting if simulated results are significant. Default is 0.05.

Value

Number of trials, out of numRuns which were detected as significant in the bootstrap testing.

References

Steinmetz, P.N. & Thorp, C.K. (2013) Testing for effects of different stimuli on neuronal firing relative to background activity. Journal of Neural Engineering, Sept. 2013.

Examples

Run this code

# 
# Compute power of changes from background test (CBT) over
# a range of plausible simulation parameters, placing results
# in a dataframe.
# This code generates results, a subset of which is shown in figure 4
# of Steinmetz & Thorp 2013.
#
## Not run: 
# require('PsumtSim')
# require('EffectsRelBaseline')
# require('boot')
# 
# if (exists('res',inherits=FALSE)) rm('res')
# 
# possibleCatsWithEffect<-c(1,2,5,10)
# numRuns<-1000
# 
# numCats in c(5,10,20,30)) {
#   catsWithEffect<-possibleCatsWithEffect[possibleCatsWithEffect<=numCats]
#   
#   for (anovaP in c(0.01,0.05)) {
#     for (bkgRate in seq(0.5,10, by=0.5)) { 
#       for (numCatsWithEffect in catsWithEffect) {
#         for (fracChange in c(0.25, 0.5, 1)) {
#         	
#           print(paste(numCats,anovaP,bkgRate,numCatsWithEffect,fracChange))
#           
#           respLevel<-bkgRate*(1+fracChange)
#       
#           numSig<-compPowerRespDetection(bkgRate,respLevel,numCats,
#                                           numCatsWithEffect,
#                                           alpha=anovaP, numRuns=numRuns,
#                                           showProgress=TRUE)
# 
#           df1<-data.frame(numCats=numCats,numCatsEffect=numCatsWithEffect,
#                         fracChange=fracChange, anovaP=anovaP, bkgRate=bkgRate,
#                         frac=numSig/numRuns)
#       
#           if (!exists('res',inherits=FALSE)) res<-df1
#           else res<-rbind(res,df1)
#         }
#       }
#     }
#   }
# }
# ## End(Not run)

#
# Example to generate short simulation corresponding to cross at 5 in figure 4B
# of Steinmetz & Thorp, 2013. Estimate here will be highly variable due to small
# number of runs.
#

require('PsumtSim')
require('EffectsRelBaseline')
require('boot')
possibleCatsWithEffect<-5
numRuns<-5
numCats<-30
catsWithEffect<-5
anovaP<-0.01
bkgRate<-1
numCatsWithEffect<-5
fracChange<-1
print(paste(numCats,anovaP,bkgRate,numCatsWithEffect,fracChange))
          
respLevel<-bkgRate*(1+fracChange)
      
numSig<-compPowerRespDetection(bkgRate,respLevel,numCats,
                                 numCatsWithEffect,
                                 alpha=anovaP, numRuns=numRuns,
                                 showProgress=TRUE)

df1<-data.frame(numCats=numCats,numCatsEffect=numCatsWithEffect,
                  fracChange=fracChange, anovaP=anovaP, bkgRate=bkgRate,
                  frac=numSig/numRuns)