[R] Finding Source of Error Message of 'Non-Unique Index Entries'

[David Winsemius]

On Jan 3, 2012, at 5:53 PM, Rich Shepard wrote:

> On Tue, 3 Jan 2012, David Winsemius wrote:
>
>> Maybe we need to backtrack a bit.
>
>  Yes. I've been trying to do this but still have too little  
> experience with
> R to be successful on my own.
>
>> You originally were complaining about an error that said you had
>> duplicated index entries as you attempted to make a zoo object. I  
>> assumed,
>> incorrectly it now appears, that you understood that an index in a  
>> zoo
>> object was a vector. You now seem to be admitting that you were  
>> trying to
>> use an entire dataframe as your index. As the acronym goes,
>> ...  DDT.
>
>  My understanding, apparently incorrect, was that read.zoo()  
> converted a
> data frame to a zoo object with the date column as the index vector.  
> From
> the read.zoo help page:
>
>    file: character string or strings giving the name of the file(s)  
> which
>          the data are to be read from/written to. ... ‘file’ can be a
>          ‘connection’ or a ‘data.frame’ (e.g., resulting from a  
> previous
>          ‘read.table’ call) that is subsequently processed to a  
> ‘"zoo"’
>          series.
>
>  Based on this, I created a subset for a single parameter:
>
> burns.tds <- subset(chemdata, stream == 'BurnsCrk', select = c(site,
> sampdate, param == 'TDS', quant), drop = T)
>
> This provides three columns:
>
>     site   sampdate quant
> 599  BC-3 1992-03-27   0.1
> 600  BC-3 1992-04-30   0.1
> 601  BC-3 1992-05-30   0.1
> 603  BC-3 1992-06-19   0.1
> 1214 BC-3 1992-07-20   0.1
> 1215 BC-3 1992-08-10   0.1
>
>  Then, to create the zoo object,
>
> burns.tds.z <- read.zoo(burns.tds, split = 1, index = 2)
>
>
>> Right. If you had said it was a dataframe, I would have suggested:
>>
>> burns.tds[ !duplicated(burns.tds) ,  ]
>>
>> But that would only identify entire duplicated rows; it would  not  
>> cure the misguided notion of creating a zoo-index from a dataframe.
>
>  Where did I go astray in trying to create a zoo-index with this  
> procedure?
> How do I extract zoo objects from a data frame?
>
>> You still have not really described what your are trying to do ...  
>> or with what data you are trying to do it with.
>
>  I have a data frame with water quality sampling data and I'm now  
> trying to
> plot time series of specific chemical concentrations (y axis) as a  
> function
> of their irregular collection over a period of 30 years or less. I  
> want to
> plot the time series for each site along the stream as a separate  
> line in
> the same panel.
>
>> You might want to think about taking that sampdate which is now a  
>> factor
>> and turinging it into a data object which would then satisfy the
>> requirements of an index for a zoo object.
>
>  I had run the read.table() function again and forgot to convert the  
> date
> column from a factor to a date. I've now done this but the result is  
> still
> the same:
>
> str(burns.tds)
> 'data.frame':	2472 obs. of  3 variables:
> $ site    : Factor w/ 137 levels "BC-0.5","BC-1",..: 5 5 5 5 5 5 5 5  
> 5 5
> $ sampdate: Date, format: "1992-03-27" "1992-04-30" ...
> $ quant   : num  0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 8.08 ...
>
> burns.tds.z <- read.zoo(burns.tds, split = 1, index = 2)
> Error in merge.zoo(C-0.5 = c(0, 0.01, 0.01, 0.06, 0.18, NA, 76.56,  :
>  series cannot be merged with non-unique index entries in a series
> In addition: Warning messages:
> 1: In zoo(rval4[[i]], ix[[i]]) :
>  some methods for “zoo” objects do not work if the index entries in
> ‘order.by’ are not unique

So NOW you need to look at :

burns.tds$sampdate[ duplicated(burns.tds$sampdate) ]

But since you are using data that is in the long format you will need  
to do it within categories determined by your site variable. I would  
be starting with tapply for that purpose, but I suppose you could just  
use

with(chamedata, table(sampdate, site) ) and see if any anomalies  
popped out.

>
>  I want to learn how to get data into zoo objects for time series  
> analyses
> so I greatly appreciate the help provided here.
>
> Rich

-- 

David Winsemius, MD
West Hartford, CT