Slideshare uses cookies to improve functionality and performance, and to provide you with relevant advertising. If you continue browsing the site, you agree to the use of cookies on this website. See our User Agreement and Privacy Policy.

Slideshare uses cookies to improve functionality and performance, and to provide you with relevant advertising. If you continue browsing the site, you agree to the use of cookies on this website. See our Privacy Policy and User Agreement for details.

Like this presentation? Why not share!

1,002 views

Published on

No Downloads

Total views

1,002

On SlideShare

0

From Embeds

0

Number of Embeds

4

Shares

0

Downloads

35

Comments

0

Likes

1

No embeds

No notes for slide

- 1. Data cleaning Stat405 Hadley Wickham Monday, 31 August 2009
- 2. 1. Intro to data cleaning 2. Missing values 3. Subsetting 4. Modifying 5. Short cuts Monday, 31 August 2009
- 3. Clean data is: Columnar (rectangular, observations in rows, variables in columns) Consistent Concise Complete Correct Monday, 31 August 2009
- 4. Correct Can’t restore correct values without original data but can remove clearly incorrect values Options: Remove entire row Mark incorrect value as missing Monday, 31 August 2009
- 5. What is a missing value? In R, written as NA. Has special behaviour: NA + 3 = ? NA > 2 = ? mean(c(2, 7, 10, NA)) = ? NA == NA ? Use is.na() to see if a value is NA Many functions have na.rm argument Monday, 31 August 2009
- 6. Your turn Look at histograms and scatterplots of x, y, z from the diamonds dataset Which values are clearly incorrect? Which values might we be able to correct? (Remember measurements are in millimetres, 1 inch = 25 mm) Monday, 31 August 2009
- 7. Plots qplot(x, data = diamonds, binwidth = 0.1) qplot(y, data = diamonds, binwidth = 0.1) qplot(z, data = diamonds, binwidth = 0.1) qplot(x, y, data = diamonds) qplot(x, z, data = diamonds) qplot(y, z, data = diamonds) Monday, 31 August 2009
- 8. Modifying data To modify, must ﬁrst know how to extract, or subset. Many different methods available in R. We’ll start with most explicit then learn some shortcuts. Basic structure: df$varname df[row index, column index] Monday, 31 August 2009
- 9. $ Remember str(diamonds) ? That hints at how to extract individual variables: diamonds$carat diamonds$price Monday, 31 August 2009
- 10. [ positive integers select speciﬁed negative integers omit speciﬁed characters extract named items nothing include everything logicals select T, omit F Monday, 31 August 2009
- 11. Challenge There is an equivalency between logical (boolean) and numerical (set) indexing. How do you change a logical index to a numeric index? And vice versa? What are the equivalents of the boolean operations for numerical indices? Monday, 31 August 2009
- 12. # Nothing str(diamonds[, ]) # Positive integers & nothing diamonds[1:6, ] # same as head(diamonds) diamonds[, 1:4] # watch out! # Positive integers * 2 diamonds[1:10, 1:4] diamonds$carat[1:100] # Negative integers diamonds[-(1:53900), -1] # Character vector diamonds[, c("depth", "table")] diamonds[1:100, "carat"] Monday, 31 August 2009
- 13. [ + logical vectors # The most complicated to understand, but # the most powerful. Lets you extract a # subset defined by some characteristic of # the data x_big <- diamonds$x > 10 head(x_big) tail(x_big) sum(x_big) diamonds$x[x_big] diamonds[x_big, ] Monday, 31 August 2009
- 14. Useful table(zeros) functions for sum(zeros) logical vectors mean(zeros) TRUE = 1; FALSE = 0 Monday, 31 August 2009
- 15. x_big <- diamonds$x > 10 diamonds[x_big, ] diamonds[x_big, "x"] diamonds[x_big, c("x", "y", "z")] small <- diamonds[diamonds$carat < 1, ] lowqual <- diamonds[diamonds$clarity %in% c("I1", "SI2", "SI1"), ] # Comparison functions: # < > <= >= != == %in% # Boolean operators small <- diamonds$carat < 1 & diamonds$price > 500 lowqual <- diamonds$colour == "D" | diamonds$cut == "Fair" Monday, 31 August 2009
- 16. And a & b Or a | b Not !b Xor xor(a, b) Monday, 31 August 2009
- 17. Saving results # Prints to screen diamonds[diamonds$x > 10, ] # Saves to new data frame big <- diamonds[diamonds$x > 10, ] # Overwrites existing data frame. Dangerous! diamonds <- diamonds[diamonds$x < 10,] Monday, 31 August 2009
- 18. diamonds <- diamonds[1, 1] diamonds # Uh oh! rm(diamonds) str(diamonds) # Phew! Monday, 31 August 2009
- 19. Your turn Extract diamonds with equal x & y. Extract diamonds with incorrect/unusual x, y, or z values. Monday, 31 August 2009
- 20. equal <- diamonds[diamonds$x == diamonds$y, ] y_big <- diamonds$y > 10 z_big <- diamonds$z > 6 x_zero <- diamonds$x == 0 y_zero <- diamonds$y == 0 z_zero <- diamonds$z == 0 zeros <- x_zero | y_zero | z_zero bad <- y_big | z_big | zeros dbad <- diamonds[bad, ] Monday, 31 August 2009
- 21. Aside: strategy The biggest problem I see new programmers make is trying to do too much at once. Break the problem into pieces and solve the smallest piece ﬁrst. Then check each piece before solving the next problem. Monday, 31 August 2009
- 22. Making new variables diamonds$pricepc <- diamonds$price / diamonds$carat diamonds$volume <- diamonds$x * diamonds$y * diamonds$z qplot(pricepc, carat, data = diamonds) qplot(carat, volume, data = diamonds) Monday, 31 August 2009
- 23. Modifying values Combination of subsetting and making new variables: diamonds$x[x_zero] <- NA diamonds$z[z_big] <- diamonds$z[z_big] / 10 These modify the data in place. Be careful! Monday, 31 August 2009
- 24. diamonds$volume <- diamonds$x * diamonds$y * diamonds$z qplot(carat, volume, data = diamonds) # Fix problems & replot diamonds$x[x_zero] <- NA diamonds$y[y_zero] <- NA diamonds$z[z_zero] <- NA diamonds$y[y_big] <- diamonds$y[y_big] / 10 diamonds$z[z_big] <- diamonds$z[z_big] / 10 diamonds$volume <- diamonds$x * diamonds$y * diamonds$z qplot(carat, volume, data = diamonds) Monday, 31 August 2009
- 25. Your turn Fix the incorrect values and replot scatterplots of x, y, and z. Are all the unusual values gone? Correct any other strange values. Hint: If qplot(a, b) is a straight line, qplot(a, a / b) will be a ﬂat line. Makes selecting strange values much easier! Monday, 31 August 2009
- 26. qplot(carat, volume, data = diamonds) qplot(carat, volume / carat, data = diamonds) weird_density <- (diamonds$volume / diamonds$carat) < 140 | (diamonds$volume / diamonds$carat) > 180 weird_density <- weird_density & !is.na(weird_density) diamonds[weird_density, c("x", "y", "z", "volume")] <- NA Monday, 31 August 2009
- 27. Short cuts You’ve been typing diamonds many many times. There are three shortcuts: with, subset and transform. These save typing, but may be a little harder to understand, and will not work in some situations. Useful tools, but don’t forget the basics. Monday, 31 August 2009
- 28. weird_density <- (diamonds$volume / diamonds$carat) < 140 | (diamonds$volume / diamonds$carat) > 180 weird_density <- with(diamonds, (volume / carat) < 140 | (volume / carat) > 180) diamonds[diamonds$carat < 1) subset(diamonds, carat < 1) equal <- diamonds[diamonds$x == diamonds$y, ] equal <- subset(diamonds, x == y) Monday, 31 August 2009
- 29. diamonds$volume <- diamonds$x * diamonds$y * diamonds$z diamonds$pricepc <- diamonds$price / diamonds$carat diamonds <- transform(diamonds, volume = x * y * z, pricepc = price / carat) Monday, 31 August 2009
- 30. Your turn Try to convert your previous statements to use with, subset and transform. Which ones convert easily? Which are hard? When is the shortcut actually a longcut? Monday, 31 August 2009
- 31. Next time Learning how to use latex: a scientiﬁc publishing program. If you’re using a laptop, please install latex from the links on the course webpage. Monday, 31 August 2009
- 32. a & b intersect(c, d) a | b union(c, d) !b setdiff(U, c) union(setdiff(c, d), xor(a, b) setdiff(d, c)) U = seq_along(a) c = which(a) a = U %in% c d = which(b) b = U %in% d Monday, 31 August 2009

No public clipboards found for this slide

×
### Save the most important slides with Clipping

Clipping is a handy way to collect and organize the most important slides from a presentation. You can keep your great finds in clipboards organized around topics.

Be the first to comment