===========

Getting and Cleaning Data's project

The full script can be found in run_analysis.R

Task 1. Merge train and test sets into a single data set (self-explained code)

After loading raw train and test data sets and merging them together, we got feature data set X.all and label vector y.all

Task 2. Extract measurements of mean and standard deviations

Following are the steps.

1. Load code.book
2. Extract mean and standard measurements by first getting their indices from codebook by grep() then extracting corresponding columns

2.1) Extract mean variables:

mean.var.codes <- grep(code.book$feature, pattern = "mean") 
mean.ds <- X.all[ ,mean.var.codes]

2.2) Here I went ahead step 4 and rename variables in mean.ds, notice that we need to pass value = T to grep() to get values instead of indices.

mean.var.names <- grep(code.book$feature, pattern = "mean", value = T)
names(mean.ds) <- mean.var.names

2.3) Proceed similarly for standard deviation variables to get std.ds 2.4) Bind mean.ds and std.ds to get first.ds

Task 3. Replace encoded class label in y.all by descriptive activity names (self-explained code)

Task 4. Label the data set by descriptive variable names (already done in step 2)

Task 5. Creates a second, independent tidy data set with the average of each variable for each activity and each subject

• Get train and test ids from train/subject_train.txt and test/subject_test.txt and bind them together to get all subject ids.
• Now I create a second data set where I pasted togethere subject ids, activity labels, mean and standard deviation variables (from first.ds)

second.ds <- data.frame("subject.id" = ids$V1, "activity" = y.all$act, stringsAsFactors = F)
second.ds <- cbind(second.ds, first.ds)

• Get average of each variable for each activity and each subject:
  • Define a function getMean(col) to get avg for a given variable for each combination of activity and subject using tapply(). Note that output of getMean() is a data frame.

getMean <- function(col) {
  tmp.res <- tapply(second.ds[ ,col], INDEX = second.ds[ ,1:2], mean)
  tmp.df <- as.data.frame(tmp.res)
  res <- data.frame("variable" = names(second.ds)[col], "subject.id" = 1:nrow(tmp.df))
  cbind(res, tmp.df)
}

• Apply the function getMean() to all columns containing mean and standard variables of second.ds (columns from 3 onwards). Here I used ldply() from package plyr (which is convenient for controlling various kinds of outputs and also has a decent option for estimating time to complete, especially important for large data sets).

require(plyr)
tidy.ds <- ldply(3:ncol(second.ds), getMean, .progress = "time")