[R] Confused: Looping in dataframes

[David Winsemius]

On Jun 25, 2010, at 7:09 AM, phani kishan wrote:

> On Fri, Jun 25, 2010 at 1:54 PM, Paul Hiemstra  
> <p.hiemstra at geo.uu.nl> wrote:
>
>> On 06/25/2010 10:02 AM, phani kishan wrote:
>>
>>> Hey,
>>> I have a data frame x which consists of say 10 vectors. I  
>>> essentially want
>>> to find out the best fit exponential smoothing for each of the  
>>> vectors.
>>>
>>> The problem while I'm getting results when i say
>>>
>>>
>>>> lapply(x,ets)
>>>>
>>>>
>>> I am getting an error when I say
>>>
>>>
>>>> myprint
>>>>>
>>>>>
>>>> function(x)
>>> {
>>> for(i in 1:length(x))
>>> {
>>> ets(x[i],model="AZZ",opt.crit=c("amse"))
>>>
>>>
>> Hi,
>>
>> Please provide a reproducible example, as stated in the posting  
>> guide. My
>> guess is that replacing x[i] by x[[i]] would solve the problem.  
>> Double
>> brackets return a vector in stead of a data.frame that has just  
>> column i.
>>
> Hey Paul,
> As requested.
> My example data frame
>
> sdata:
>                SKU1            SKU2           SKU3           SKU4
> 1               583.8             574.6          1106.9            
> 648.1
> 2               441.7             552.8          1021.3            
> 353.6
> 3               454.2             555.7           998.3            
> 306.4
> 4               569.7             507.6           811.1            
> 360.7
> 5               512.3             620.0          1046.3            
> 713.9
> 6               580.8             668.2           732.0            
> 490.9
> 7               648.5             766.9           653.4            
> 422.1
> 8               617.4             657.1           602.1            
> 190.8
> 9               826.8             767.3           640.5            
> 324.1
> 10             1163.0             657.6           429.6            
> 181.1
> 11              643.5             788.9           569.1            
> 331.9
> 12              846.9             568.6           425.1            
> 224.6
> 13              580.7             582.9           434.2            
> 226.9
>
> now when I apply
> lapply(sdata,ets)
> I get a result as:
> $SKU1
> ETS(A,N,N)
>
> Call:
> ets(y = x, model = "AZZ")
>
>  Smoothing parameters:
>    alpha = 0.3845
>
>  Initial states:
>    l = 533.3698
>
>  sigma:  181.7615
>
>     AIC     AICc      BIC
> 172.6144 173.8144 173.7443
>
> $SKU2
> ETS(A,N,N)
>
> Call:
> ets(y = x, model = "AZZ")
>
>  Smoothing parameters:
>    alpha = 0.5026
>
>  Initial states:
>    l = 567.821
>
>  sigma:  86.7074
>
>     AIC     AICc      BIC
> 153.3704 154.5704 154.5003
>
> $SKU3
> ETS(A,A,N)
>
> Call:
> ets(y = x, model = "AZZ")
>
>  Smoothing parameters:
>    alpha = 1e-04
>    beta  = 1e-04
>
>  Initial states:
>    l = 1189.2221
>    b = -64.3776
>
>  sigma:  85.4153
>
>     AIC     AICc      BIC
> 156.9800 161.9800 159.2398
>
> $SKU4
> ETS(A,A,N)
>
> Call:
> ets(y = x, model = "AZZ")
>
>  Smoothing parameters:
>    alpha = 1e-04
>    beta  = 1e-04
>
>  Initial states:
>    l = 566.9001
>    b = -27.8818
>
>  sigma:  127.2654
>
>     AIC     AICc      BIC
> 167.3475 172.3475 169.6073
>
> Now when I run the same using:
> myfun<-function(x)
> {
> for(i in 1:length(x))
> {
> ets(x[i])
>>
>> }
> }
> I got the error as mentioned before. Now on modifying it to
> myfun<-function(x)
> {
> for(i in 1:length(x))
> {
> return(ets(x[[i]])
> }
> }
> I only got the output as
> ETS(A,N,N)
>
> Call:
> ets(y = x[[i]], model = "AZZ", opt.crit = c("amse"))
>
>  Smoothing parameters:
>    alpha = 0.3983
>
>  Initial states:
>    l = 516.188
>
>  sigma:  181.8688
>
>     AIC     AICc      BIC
> 172.6298 173.8298 173.7597
>
> I think its considering whole dataframe as a series.

Doubtful. It is quietly calculating all of the requested models but  
you did not do anything with them inside the loop (which is a  
function). You could have assigned them to something permanent or  
printed them (or both):

ets_x <- list()
> for(i in 1:length(x))
> {
> print(ets(x[[i]]); ets_x <- c(ets_x, ets(x[[i]])
> }
> }

ets_x

> As said my objective it to essentially come up with a best  
> exponential model
> for each of the SKU's in the dataframe. However I want to be able to  
> extract
> information like mse, mape etc later. So kindly suggest.
>
> Thanks in advance,
> Phani
>
>