[R] Complicated analysis for huge databases

[Boris Steipe]

Something like the following?

AllMAFs <- list()

for (i in length(SeparatedGroupsofmealsCombs) {
  AllMAFs[[i]] <- apply(SeparatedGroupsofmealsCombs[[i]], 2, function(x)maf(tabulate(x+1)))
}


(untested, of course)
Also the solution is a bit generic since I don't know what the output of maf() looks like in your case, and I don't understand why you use tabulate because I would have assumed that's what maf() does - but that's not for me to worry about :-) 



B.



> On Nov 17, 2017, at 7:15 PM, Allaisone 1 <allaisone1 at hotmail.com> wrote:
> 
> 
> Thanks Boris , this was very helpful but I'm struggling with the last part.
> 
> 1) I combined the first 2 columns :-
> 
>  
> library(tidyr)
> SingleMealsCode <-unite(MyData, MealsCombinations, c(MealA, MealB), remove=FALSE)
> SingleMealsCode <- SingleMealsCode[,-2]
> 
>   2) I separated this dataframe into different dataframes based on "MealsCombination"
>    column so R will recognize each meal combination separately :
> 
> SeparatedGroupsofmealsCombs <- split(SingleMealCode,SingleMealCode$MealsCombinations)
> 
> after investigating the structure of "SeparatedGroupsofmealsCombs" , I can see 
> a list of different databases, each of which represents a different Meal combinations which is great.
> 
> No, I'm struggling with the last part, how can I run the maf code for all dataframes?
> 
> when I run this code as before :-
> 
> maf <- apply(SeparatedGroupsofmealsCombs, 2, function(x)maf(tabulate(x+1))) 
> 
> an error message says : dim(X) must have a positive length . I'm not sure which length 
> I need to specify.. any suggestions to correct this syntax ?  
> 
> Regards
> Allaisone
> From: Boris Steipe <boris.steipe at utoronto.ca>
> Sent: 17 November 2017 21:12:06
> To: Allaisone 1
> Cc: R-help
> Subject: Re: [R] Complicated analysis for huge databases
>  
> Combine columns 1 and 2 into a column with a single ID like "33.55", "44.66" and use split() on these IDs to break up your dataset. Iterate over the list of data frames split() returns. 
> 
> 
> B.
> 
> > On Nov 17, 2017, at 12:59 PM, Allaisone 1 <allaisone1 at hotmail.com> wrote:
> > 
> > 
> > Hi all ..,
> > 
> > 
> > I have a large dataset of around 600,000 rows and 600 columns. The first col is codes for Meal A, the second columns is codes for Meal B. The third column is customers IDs where each customer had a combination of meals. Each column of the rest columns contains values 0,1,or 2. The dataset is organised in a way so that the first group of customers had similar meals combinations, this is followed by another group of customers with similar meals combinations but different from the first group and so on. The dataset looks like this :-
> > 
> > 
> >> MyData
> > 
> >       Meal A     Meal B     Cust.ID      I            II        III     IV   ...... 600
> > 
> > 1    33                 55             1             0           1        2       0
> > 
> > 2    33                 55              3             1          0        2        2
> > 
> > 3    33                 55              5             2          1        1         2
> > 
> > 4    44                 66               7            0          2         2        2
> > 
> > 5   44                  66               4            1          1          0       1
> > 
> > 6   44                  66                9            2          0          1       2
> > 
> > .
> > 
> > .
> > 
> > 600,000
> > 
> > 
> > 
> > I wanted to find maf() for each column(from 4 to 600) after calculating the frequency of the 3 values (0,1,2) but this should be done group by group (i.e. group(33-55) : rows 1:3 then group(44-66) :rows 4:6 and so on).
> > 
> > 
> > I can do the analysis  for the entire column but not group by group like this :
> > 
> > 
> > MAF <- apply(MyData[,4:600], 2, function(x)maf(tabulate(x+1)))
> > 
> > How can I modify this code to tell R to do the analysis group by group for each column so I get maf value for 33-55 group of clolumn I, then maf value for group 44-66 in the same column I,then the rest of groups in this column and do the same for the remaining columns.
> > 
> > In fact, I'm interested in doing this analysis for only 300 columns but all of the 600 columns.
> > I have another sheet contains names of columns of interest like this :
> > 
> >> ColOfinterest
> > 
> > Col
> > I
> > IV
> > V
> > .
> > .
> > 300
> > 
> > Any one would help with the best combination of syntax to perform this complex analysis?
> > 
> > Regards
> > Allaisone
> > 
> > 
> > 
> > 
> > 
> > 
> > 
> >        [[alternative HTML version deleted]]
> > 
> > ______________________________________________
> > R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> > https://stat.ethz.ch/mailman/listinfo/r-help
> > PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> > and provide commented, minimal, self-contained, reproducible code.