Question

亲爱的用户我正在研究重新采样我的时间序列数据的方法。基本上我的问题是我必须使用具有不同时间戳的不同信号进行计算。一开始我想用aproxfun（）或者动物园包。

问题是我需要保留t-1和t + 1的NAN值，以防它们存在且只有在满足特定条件的情况下。为了提取信号的不同时间戳，我做了以下一个：

GlobalTime<-sort(do.call(`c`, data[c(1, 3, 5)])) #extract the total timestamp of the measurements
Time1<-seq.POSIXt(from=min(data$X1, na.rm=TRUE), to=max(data$X1, na.rm=TRUE), by="s") #extract the timestamp of each measurment
Time2<-seq.POSIXt(from=min(data$X3, na.rm=TRUE), to=max(data$X3, na.rm=TRUE), by="s")
Time3<-seq.POSIXt(from=min(data$X5, na.rm=TRUE), to=max(data$X5, na.rm=TRUE), by="s")
#All vectors have different size

对于插值，我需要考虑3点。第一个，来自Time1，另外两个来自Global Time（这两个是最接近Time1 [i]的）。我认为这可以这样做：

upperlimits1=c()
lowerlimits1=c()
for (i in 1:length(actualTime)) {
    upperlimits1[i]=min(Mod(actualTime[i+1]-requestTime1[i]))
    lowerlimits1[i]=min(Mod(actualTime[i-1]-requestTime1[i]))     
}

稍后为了使插值做类似的事情：

for (i in 1:length(actualTime)) {

if(lowerlimits1[i]==Time1[i])
resample[i,1]<-value[Time[i-1]]
resample[i,2]<-Time1[i-1]
resample[i,3]<-"OK"
i++    
} else if(upperlimits1[i]==Time1[i]){
resample[i,1]<-value[Time[i+1]]
resample[i,2]<-Time1[i+1]
resample[i,3]<-"OK"
i++    
} else if (upperlimits1[i]!=Time1[i] & lowerlimits1[i]!=Time1[i] & aux[i-1]=="Wrong"){
resample[i,1]<-value[Time1[i-1]]
resample[i,2]<-Time1[i-1]
resample[i,3]<-"Wrong"
i++
} else if (upperlimits1[i]!=Time1[i] & lowerlimits1[i]!=Time1[i] & aux[i+1]=="Wrong"){
resample[i,1]<-value[Time1[i+1]]
resample[i,2]<-Time1[i+1]
resample[i,3]<-"Wrong"
i++
} else if (aux[i]=="C"){
resample<-value[Time1[i-1]]
resample[i,2]<-Time1[i-1]
resample[i,3]<-aux[i-1]
i++
} else {
Delta=upperlimits1[i]-lowerlimits1[i]
if (Delta>3600){
resample[i,3]<-"Wrong"
} if (Delta<3600){
resample[i,3]<-"OK"
coeff=(Time1[i]-lowerlimits1[i]) / (upperlimits1[i]-lowerlimits1[i]);
resample[i,1]=value[lowerlimits1[i]] + (value[upperlimits1[i] - value[lowerlimits1[i]]) * coeff);
i++
}

编辑：根据要求，我添加了数据的样子：

Time1    Value1   Quality1   Time2      Value2    Quality2...
00:00.9  41.3       Ok       00:04.0    78.2         Ok                               
00:01.9  41.68      Ok       00:07.0    78.5         Ok                                
00:04.9  35.34      Ok       00:08.0    66           Ok                                 
00:15.9  35.98      Ok       00:14.0    65.8         Ok                                 
00:21.9  Wrong      Wrong    00:15.0    64.5         Ok                                   
00:22.9  38.9       Ok       00:19.0    40.5         Ok                                  
59:56.5  40.1       Ok       00:21.0    30.5         Ok
#of course sometimes wrong values can be recorded for longer periods as hours or days
#at the end the interpolation has the purpose of having the three or more signals with same timestamp

所需的输出

Time1         Value1       Quality1   Time2      Value2    Quality2...
    01:00:00  IntValue       Ok       01:00:00   IntValue      Ok                               
    02:00:00  IntValue       Ok       02:00:00   IntValue      Ok                                
    03:00:00  IntValue       Ok       03:00:00   IntValue      Ok                                 
    04:00:00  IntValue       Ok       04:00:00   IntValue      Ok                                 
    05:00:00  IntValue       Ok       05:00:00   IntValue      Ok                                   
    06:00:00  IntValue       Ok       06:00:00   IntValue      Ok                                  
    07:00:00  IntValue       Ok       07:00:00   IntValue      Ok

但是如果质量列错误超过24小时，插值应写入

然而，您可以想象代码不起作用。有人试过这个吗？

BR

R中的自定义重采样方法

0 个答案: