Using a column that contains a frequency/weight/count in R [closed] - r

Question

Closed. This question needs details or clarity. It is not currently accepting answers.
Want to improve this question? Add details and clarify the problem by editing this post.
Closed 3 years ago.
Improve this question
This is an easy question to ask, but a hard one to search for. Frequency is used all over the place. I tried a synonym (weight), but since mtcars is so widely used, I get a lot of false negatives as well. Same thing for counts.
I'm looking at datasets::HairEyeColor, partly reproduced here:
Hair Eye Sex Freq
1 Black Brown Male 32
2 Brown Brown Male 53
3 Red Brown Male 10
4 Blond Brown Male 3
5 Black Blue Male 11
6 Brown Blue Male 50
7 Red Blue Male 10
8 Blond Blue Male 30
9 Black Hazel Male 10
10 Brown Hazel Male 25
.
.
.
I can across this when trying to show someone how to make a mosaic plot of any two of Hair, Eye, and Gender. On first read, I didn't see a way to specify a column to specify "this column represents 32 of the set members" but I didn't read too carefully.
I suppose I could reshape the data using melt() and reshape() every time I receive data with a frequency column, but that seems kind of drastic.
In other languages I know, I could add a parameter to the fitting function to let it know “there’s not just one row with this set of levels, there are n of them. So if I wanted to see a distribution, I might say
DISTR(Y=Hair, FREQ=freq)
...which would generate a histogram or density plot with n values per row
Alternately,
lm(hair ~ eye + sex, data = ‘HairEyeColor’, freq = ‘freq’)
Would fit a linear model with 32 replications if the first row rather than 1.
I’m asking about a way to use the 32 in the first row (for example) to tell the modeling or graphing function that there are 32 cases with this combination of levels, 53 with the combination in the second row, etc.
Surely this kind of data shows up a lot. I see it all the time, but there’s usually a way to say that this number specifies the frequency that this row represents in the actual data. Rather than a data table with 32 rows of Black, Brown, Male, there’s one row with frequency 32.
(No plyr please.)

Answer 1

No, there is not a standard way to use this type of data across all of R.
Many of the basic modeling functions, e.g., lm, glm, nls, loess, and more from the stats package accept a weights argument that will meet your needs. prop.test accepts data in either format. But many other modeling functions do not, e.g., knn, princomp, and many others not in base R.
barplot accepts input in either format. mosaicplot expects input as an aggregated contingency table. Other types of plots would require more custom handling, because there are a lot of different things you could do with frequency.
Of course, anything not in base R is up to whoever writes it.
ggplot2 (which is not base R) generally handles this really well, e.g., geom_bar will stack up values by default, or in the case of scatterplots you could map size or color or alpha to visually convey the intensity.
randomForest and xgboost do not accept weights
I will say that I very rarely find this to be a problem. I'd encourage you to ask specific questions about methods where it is causing you issues. I think mosaicplot is a bad example as it expects a contingency table, so the problem would be the opposite: using it with disaggregated data would require first aggregating it up to a frequency table.