Hello All, Eric Langley here with my first post to this list. I am looking to determine if R is suitable for a development project I am working on and if so possibly finding someone proficient in R that would be interested in doing the coding. I would like to preface my inquiry that while I am not a programmer I can communicate in a dialog my objectives. An array of rank ordered data looks like this: Item-Rank First Second Third Fourth Totals Item1 6 8 0 0 14 Item2 7 5 2 0 14 Item3 1 1 11 1 14 Item4 0 0 1 13 14 Totals 14 14 14 14 The required output of R will be two fold; 1, a numerical score for each of the Items (1-4) from highest to lowest and lowest to highest on a scale of 0-99 that is statistically accurate. For this example the scores would be Item1 highest number down to Item4 with the lowest number. In reverse Item4 would be the highest number down to Item1 the lowest number. For the Highest like this; Item1=94, Item2=88, Item3=48, Item4=2 (just guessing here on the scores...:) 2, a graphical output of the data based on the scores in three special graphs with a middle line at '0' and increasing numbers to the left AND right. The graphs plot the Highest ranked Items, the Lowest Ranked items and a combination of the two. Sample graphs are here: http://community.abeo.us/sample-graphs/ Looking forward to hearing if R will be able to accomplish this. TYIA, ~eric -- Eric Langley Founder abeo eric at abeo.us 404-326-5382
R suitability for development project
7 messages · Rolf Turner, Eric Langley, Jim Lemon
On 03/09/12 14:15, Eric Langley wrote:
Hello All, Eric Langley here with my first post to this list. I am looking to determine if R is suitable for a development project I am working on and if so possibly finding someone proficient in R that would be interested in doing the coding.
Is there monetary reward involved? Or are you just counting
on people's good nature?
I would like to preface my inquiry that while I am not a programmer I can communicate in a dialog my objectives. An array of rank ordered data looks like this: Item-Rank First Second Third Fourth Totals Item1 6 8 0 0 14 Item2 7 5 2 0 14 Item3 1 1 11 1 14 Item4 0 0 1 13 14 Totals 14 14 14 14 The required output of R will be two fold; 1, a numerical score for each of the Items (1-4) from highest to lowest and lowest to highest on a scale of 0-99 that is statistically accurate. For this example the scores would be Item1 highest number down to Item4 with the lowest number. In reverse Item4 would be the highest number down to Item1 the lowest number. For the Highest like this; Item1=94, Item2=88, Item3=48, Item4=2 (just guessing here on the scores...:) 2, a graphical output of the data based on the scores in three special graphs with a middle line at '0' and increasing numbers to the left AND right. The graphs plot the Highest ranked Items, the Lowest Ranked items and a combination of the two. Sample graphs are here: http://community.abeo.us/sample-graphs/ Looking forward to hearing if R will be able to accomplish this.
This reminds me of
fortune("driveway")
from the "fortunes" package.
cheers,
Rolf Turner
Eric Langley wrote:
Hello All, Eric Langley here with my first post to this list. I am looking to determine if R is suitable for a development project I am working on and if so possibly finding someone proficient in R that would be interested in doing the coding. I would like to preface my inquiry that while I am not a programmer I can communicate in a dialog my objectives. An array of rank ordered data looks like this: Item-Rank First Second Third Fourth Totals Item1 6 8 0 0 14 Item2 7 5 2 0 14 Item3 1 1 11 1 14 Item4 0 0 1 13 14 Totals 14 14 14 14 The required output of R will be two fold; 1, a numerical score for each of the Items (1-4) from highest to lowest and lowest to highest on a scale of 0-99 that is statistically accurate. For this example the scores would be Item1 highest number down to Item4 with the lowest number. In reverse Item4 would be the highest number down to Item1 the lowest number. For the Highest like this; Item1=94, Item2=88, Item3=48, Item4=2 (just guessing here on the scores...:) 2, a graphical output of the data based on the scores in three special graphs with a middle line at '0' and increasing numbers to the left AND right. The graphs plot the Highest ranked Items, the Lowest Ranked items and a combination of the two. Sample graphs are here: http://community.abeo.us/sample-graphs/ Looking forward to hearing if R will be able to accomplish this.
Hi Eric,
I would use mean ranks for something like this. You would have to
calculate these from your summary array unless you have the raw ranks.
ranksumm2meanranks<-function(x,nobs) {
nitems<-dim(x)[1] - 1
meanrankvec<-rep(0,nitems)
for(rankrow in 1:nitems) {
for(rankcol in 1:nitems)
meanrankvec[rankrow]<-
meanrankvec[rankrow]+x[rankrow,rankcol]*rankcol
meanrankvec[rankrow]<-
meanrankvec[rankrow]/x[rankrow,nitems+1]
}
names(meanrankvec)<-rownames(x)[-1]
return(meanrankvec)
}
> ranksumm2meanranks(x)
Item2 Item3 Item4 Totals
1.571429 1.642857 2.857143 3.928571
You can obtain the "reversed" ranks by subtracting the above from the
maximum rank score (4), but I don't see why you would want to do this.
Your explanation of the plot is not entirely clear. The ranges of the
ranks for the items are:
Item1 c(1,2)
Item2 c(1,3)
Item3 c(1,4)
Item4 c(3,4)
You could plot these as horizontal bars spanning the range of the ranks
for each item with a vertical line across each bar showing the value of
the mean rank for that item. This would illustrate both the relative
position and variability of ranks, something like a boxplot.
In case you have incomplete ranks, check the crank package for
completion of incomplete ranks.
Jim
Rolf wrote:
Is there monetary reward involved? Or are you just counting
on people's good nature?
I note: Absolutely, I should have noted that...
Rolf wrote:
This reminds me of
fortune("driveway")
from the "fortunes" package.
I note: You mean this one? "I think this is kind of like asking ?will your Land Rover make it up my driveway?", but I?ll assume the question was asked in all seriousness. ?Ista Zahn (in response to a request for replication of some data preprocessing done in SAS) R-help (April 2011)"
On Mon, Sep 3, 2012 at 2:08 AM, Rolf Turner <rolf.turner at xtra.co.nz> wrote:
On 03/09/12 14:15, Eric Langley wrote:
Hello All, Eric Langley here with my first post to this list. I am looking to determine if R is suitable for a development project I am working on and if so possibly finding someone proficient in R that would be interested in doing the coding.
Is there monetary reward involved? Or are you just counting
on people's good nature?
I would like to preface my inquiry that while I am not a programmer I can communicate in a dialog my objectives. An array of rank ordered data looks like this: Item-Rank First Second Third Fourth Totals Item1 6 8 0 0 14 Item2 7 5 2 0 14 Item3 1 1 11 1 14 Item4 0 0 1 13 14 Totals 14 14 14 14 The required output of R will be two fold; 1, a numerical score for each of the Items (1-4) from highest to lowest and lowest to highest on a scale of 0-99 that is statistically accurate. For this example the scores would be Item1 highest number down to Item4 with the lowest number. In reverse Item4 would be the highest number down to Item1 the lowest number. For the Highest like this; Item1=94, Item2=88, Item3=48, Item4=2 (just guessing here on the scores...:) 2, a graphical output of the data based on the scores in three special graphs with a middle line at '0' and increasing numbers to the left AND right. The graphs plot the Highest ranked Items, the Lowest Ranked items and a combination of the two. Sample graphs are here: http://community.abeo.us/sample-graphs/ Looking forward to hearing if R will be able to accomplish this.
This reminds me of
fortune("driveway")
from the "fortunes" package.
cheers,
Rolf Turner
-- Eric Langley Founder eric at abeo.us 404-326-5382
Jim wrote:
I would use mean ranks for something like this. You would have to calculate these from your summary array unless you have the raw ranks.
I note: Thank you for the detailed answer. I have the raw ranks. For a question they look like this; 1=Item2, 2=Item1, 3=Item3, 4=Item4
Jim wrote:
ranksumm2meanranks<-function(x,nobs) {
nitems<-dim(x)[1] - 1
meanrankvec<-rep(0,nitems)
for(rankrow in 1:nitems) {
for(rankcol in 1:nitems)
meanrankvec[rankrow]<-
meanrankvec[rankrow]+x[rankrow,rankcol]*rankcol
meanrankvec[rankrow]<-
meanrankvec[rankrow]/x[rankrow,nitems+1]
}
names(meanrankvec)<-rownames(x)[-1]
return(meanrankvec)
}
ranksumm2meanranks(x)
Item2 Item3 Item4 Totals 1.571429 1.642857 2.857143 3.928571
I note: Is it possible to have the output as an integer where 99 is the highest score?
Jim wrote:
Your explanation of the plot is not entirely clear. The ranges of the ranks for the items are: Item1 c(1,2) Item2 c(1,3) Item3 c(1,4) Item4 c(3,4) You could plot these as horizontal bars spanning the range of the ranks for each item with a vertical line across each bar showing the value of the mean rank for that item. This would illustrate both the relative position and variability of ranks, something like a boxplot. In case you have incomplete ranks, check the crank package for completion of incomplete ranks.
I note: The look I am aiming to achieve (as shown here: http://community.abeo.us/sample-graphs/ ) is a relative position within the middle zero based horizontal axis. The mean is not required. Since all bars are 14 units long the upper and lower values note where the end of each bar should align, either to the right for Highest or to the left for Lowest. The third graph shows both. ~eric
On Mon, Sep 3, 2012 at 7:41 AM, Jim Lemon <jim at bitwrit.com.au> wrote:
Eric Langley wrote:
An array of rank ordered data looks like this: Item-Rank First Second Third Fourth Totals Item1 6 8 0 0 14 Item2 7 5 2 0 14 Item3 1 1 11 1 14 Item4 0 0 1 13 14 Totals 14 14 14 14 The required output of R will be two fold; 1, a numerical score for each of the Items (1-4) from highest to lowest and lowest to highest on a scale of 0-99 that is statistically accurate. For this example the scores would be Item1 highest number down to Item4 with the lowest number. In reverse Item4 would be the highest number down to Item1 the lowest number. For the Highest like this; Item1=94, Item2=88, Item3=48, Item4=2 (just guessing here on the scores...:) 2, a graphical output of the data based on the scores in three special graphs with a middle line at '0' and increasing numbers to the left AND right. The graphs plot the Highest ranked Items, the Lowest Ranked items and a combination of the two. Sample graphs are here: http://community.abeo.us/sample-graphs/ Looking forward to hearing if R will be able to accomplish this.
Eric Langley Founder eric at abeo.us 404-326-5382
On 09/03/2012 11:18 PM, Eric Langley wrote:
<chomp>
I note: Is it possible to have the output as an integer where 99 is the highest score?
It certainly is. The mean ranks are:
Item2 Item3 Item4 Totals
1.571429 1.642857 2.857143 3.928571
To make the highest score 99 (meaning the score for the highest ranked
item) and the lowest score -99 as on your examples:
# first convert high ranks to high numeric scores
revmeanranks<-4-meanranks
revmeanranks
Item1 Item2 Item3 Item4
2.428571 2.357143 1.142857 0.071429
# scale the score to the desired range
library(plotrix)
rescale(revmeanranks,c(-99,99))
Item1 Item2 Item3 Item4
99.000000 93.000046 -9.000015 -99.000000
# then just round or truncate your values to get integers
<chomp>
I note: The look I am aiming to achieve (as shown here: http://community.abeo.us/sample-graphs/ ) is a relative position within the middle zero based horizontal axis. The mean is not required. Since all bars are 14 units long the upper and lower values note where the end of each bar should align, either to the right for Highest or to the left for Lowest. The third graph shows both. ~eric
I think you may be doing something here that you don't intend. The point of plots like this is to transform numeric values into lengths or areas. If you don't maintain the same metric throughout the plot, the relationship between the two is lost. If you were using vertical lines to indicate the transformed mean ranks of the items, they could be placed at the appropriate positions. As you use different edges of the bars to place them, let's see, your bars are about 90 units wide so there would be about a 180 unit offset between positive and negative transformed mean ranks. Are you sure that you want to do this? Jim
Jim, Thank you again for your response. The integer conversion looks good.
You wrote:
I think you may be doing something here that you don't intend. The point of plots like this is to transform numeric values into lengths or areas. If you don't maintain the same metric throughout the plot, the relationship between the two is lost. If you were using vertical lines to indicate the transformed mean ranks of the items, they could be placed at the appropriate positions. As you use different edges of the bars to place them, let's see, your bars are about 90 units wide so there would be about a 180 unit offset between positive and negative transformed mean ranks. Are you sure that you want to do this?
I note: Yes, the bars are half the width of the axis. In this model there are 14 questions to determine the ranking. My concept is the all of the Items are ranked. Even though some Items may not be ranked in some positions, 1,2,4 or 4, they still receive a rank. Therefore with 14 ranks all of the bars are the same width. There position along the horizontal axis is determined by the right edge of the bar for highest rank. ~eric
On Tue, Sep 4, 2012 at 5:31 AM, Jim Lemon <jim at bitwrit.com.au> wrote:
On 09/03/2012 11:18 PM, Eric Langley wrote: <chomp>
I note: Is it possible to have the output as an integer where 99 is the highest score?
It certainly is. The mean ranks are:
Item2 Item3 Item4 Totals
1.571429 1.642857 2.857143 3.928571
To make the highest score 99 (meaning the score for the highest ranked item)
and the lowest score -99 as on your examples:
# first convert high ranks to high numeric scores
revmeanranks<-4-meanranks
revmeanranks
Item1 Item2 Item3 Item4
2.428571 2.357143 1.142857 0.071429
# scale the score to the desired range
library(plotrix)
rescale(revmeanranks,c(-99,99))
Item1 Item2 Item3 Item4
99.000000 93.000046 -9.000015 -99.000000
# then just round or truncate your values to get integers
<chomp>
I note: The look I am aiming to achieve (as shown here: http://community.abeo.us/sample-graphs/ ) is a relative position within the middle zero based horizontal axis. The mean is not required. Since all bars are 14 units long the upper and lower values note where the end of each bar should align, either to the right for Highest or to the left for Lowest. The third graph shows both. ~eric
I think you may be doing something here that you don't intend. The point of plots like this is to transform numeric values into lengths or areas. If you don't maintain the same metric throughout the plot, the relationship between the two is lost. If you were using vertical lines to indicate the transformed mean ranks of the items, they could be placed at the appropriate positions. As you use different edges of the bars to place them, let's see, your bars are about 90 units wide so there would be about a 180 unit offset between positive and negative transformed mean ranks. Are you sure that you want to do this? Jim
Eric Langley Founder eric at abeo.us 404-326-5382