![Use the approxfun()function to graph the
interpolation lines.
Use similar categorical information as used for the plot function, and first try the “linear” option. You can
use “2” as a rule to treat the graph as though it was continuing through the points on the ends. The Category is
your vector with y-values. Note if you have your data in columns, you will use Category[,1] instead. f<-
approxfun(seq(start,end,1),Category[1,],method="linear", rule = 2) Start = 1997
and end = 2017 in this case. You then need to use the curve()function and make sure you specify add =
TRUE. I used purple as the color below.](https://image.slidesharecdn.com/graphinginterpolationandregressioninr-190307040629/85/How-to-combine-interpolation-and-regression-graphs-in-R-7-320.jpg)
![Use the constant argument of the
approxfun()function to see trends.
Use the same categorical information as done before and now use the “constant” option. You can use “2” as
an option to treat the graph as though it was continuing through the points.
f<-approxfun(seq(start,end,1),Category[1,],method=“constant", rule = 2)
Start and end would again be your beginning and ending years. You then need to use the curve function and
make sure you specify add = TRUE. I used lightgreen as the color below.](https://image.slidesharecdn.com/graphinginterpolationandregressioninr-190307040629/85/How-to-combine-interpolation-and-regression-graphs-in-R-8-320.jpg)
![Use the lm() curves to set up your
polynomial graphs
• Set up your regression picking the degree you guessed from the previous page:
r<-lm(Category[1,]~poly(seq(start,end,1),4,raw = T))
• Now use the predict function to get a regression function out of this:
rc<-predict(r)](https://image.slidesharecdn.com/graphinginterpolationandregressioninr-190307040629/85/How-to-combine-interpolation-and-regression-graphs-in-R-10-320.jpg)
How to combine interpolation and regression graphs in R
- 1. Graphing interpolation, regression, and subsets of Census data in R by Doug Loqa Pull data into R, group data subsets, and create a comparison between interpolation and regression curves.
- 2. You may start with the groups on the left, but prefer to summarize some condensed categories like those on the right. Measures of income dispersion 2017 2016 ….. 10th percentile limit 14,219 13,901 ….. 20th percentile limit 24,638 24,518 ….. 40th percentile limit 47,110 46,581 ….. 50th (median) 61,372 60,309 ….. 60th percentile limit 77,552 76,479 ….. 80th percentile limit 126,855 123,621 ….. 90th percentile limit 179,077 174,203 ….. 95th percentile limit 237,034 230,095 ….. Measures of income dispersion 2017 2016 ….. 50th percentile below 36834.75 36327.25 ….. Upper-Middle 102,204 100,050 ….. Top 10 % 208,056 202,149 …..
- 3. Save the file from Census data, and save the tabular data you want into a .csv. Then import it into R (The categorical data will read horizontally seen below). Use the read.csv command and make sure headers are set to true, and StringsAsFactors are set to false. The StringsAsFactors can cause information to be “bucketed” and make analysis more difficult. read.csv(“Your data.csv”, headers = T, StringsAsFactors = F)
- 4. Convert the input to numerical data for R Combine the gsub and as.numeric() functions gsub(“,”,””,datarow) as.numeric(datavector) combine below as.numeric(gsub(…..)) Assign this to a variable Problems: The data will come in as strings, so you won’t be able to run any calculations initially. Also, R doesn’t work with commas when computing values. Solutions: Run the formulas to the left to fix this. The combination of the two functions create numerical data vectors R can read.
- 5. Create all numeric vectors needed for your first category Run the mean function inside the apply() function with your saved vectors apply(rbind(vect1, vect2, vect3),2,mean) In the second argument here, you can enter a “1” or “2” to calculate across rows, or down columns respectively. In this case, since we need to calculate down, a 2 was entered. Remember this was entered horizontally. Continue this process until you have all of your categorial rows of data to analyze. Problem: You still need to combine vectors for your aggregated categorical information. Solution: The way you can summarize the vectors of data is by using the apply() function. You need to know what direction the data will be computed in, so pay attention to the second parameter here.
- 6. Use the plot()function with specified titles, colors, and the category. Use, col =“color”, xlab, ylab, main, type =“p”, and you can use seq(1997,2017,1) for x-values here. You can use your categorical data for your y-values. The x-values should represent the span of data you are measuring
- 7. Use the approxfun()function to graph the interpolation lines. Use similar categorical information as used for the plot function, and first try the “linear” option. You can use “2” as a rule to treat the graph as though it was continuing through the points on the ends. The Category is your vector with y-values. Note if you have your data in columns, you will use Category[,1] instead. f<- approxfun(seq(start,end,1),Category[1,],method="linear", rule = 2) Start = 1997 and end = 2017 in this case. You then need to use the curve()function and make sure you specify add = TRUE. I used purple as the color below.
- 8. Use the constant argument of the approxfun()function to see trends. Use the same categorical information as done before and now use the “constant” option. You can use “2” as an option to treat the graph as though it was continuing through the points. f<-approxfun(seq(start,end,1),Category[1,],method=“constant", rule = 2) Start and end would again be your beginning and ending years. You then need to use the curve function and make sure you specify add = TRUE. I used lightgreen as the color below.
- 9. Use the interpolation curves to determine what degree of polynomial to use Where it is obvious, you will see the “angled” interpolations switch from fitting above the linear to below the linear, and vice versa. Count how many of those “shifts” there are and that is your degree. So, counting 5 arrows below means there are 5-1 = 4 shifts. You can guess this requires a 4th degree polynomial.
- 10. Use the lm() curves to set up your polynomial graphs • Set up your regression picking the degree you guessed from the previous page: r<-lm(Category[1,]~poly(seq(start,end,1),4,raw = T)) • Now use the predict function to get a regression function out of this: rc<-predict(r)
- 11. Use the lines() function to graph your regression curve on top of this. You use the same sequence using the beginning and ending years, and use the regression curve based on the predict() function, generated in last step, to graph this. lines(y~x, col=“color choice”,lwd = 2)
- 12. Use the par()function to group your graphs once you have a number of them complete. par(mfrow = c(1,2)) will give you 2 graphs side-by-side. Make sure that your titles are short enough to not collide when graphing.
- 13. Add rows of graphs and you’re done! You can make this process even easier by combining some of these steps into a user- defined function if you know how to set up your variables. Also consider using the layout() function instead of par()