R Tutorial: Describing survey results

Want to learn more? Take the full course at https://learn.datacamp.com/courses/survey-and-measurement-development-in-r at your own pace. More than a video, you'll learn hands-on coding & quickly apply skills to your daily work. --- Like with most statistics, we need to check that basic assumptions about our survey data are met before moving on with the analysis. Let's look at three such sets of statistics. First, here are some rules of thumb on dealing with missing values. If there's fewer than five percent missing values in a dataset, and those values are equally dispersed across all variables, then the safest and easiest approach is to dismiss those cases. If there are more, it's back to item generation. We can get a sense of the missingness of our dataset by checking the number of rows containing a missing case, and a count of missing records by item. We'll do this by using some base R functions which count the records equal to NA. If missingness is a problem, we can plot a hierarchical clustering of the missing values with naclus() from the Hmisc package. While the statistics behind the plot are beyond the scope of this course, what's important to understand is that it shows which items tend to be missing in the same row. Finally, item correlations. We want to find groups of items that correlate highly with each other, but not so highly with items outside their group. We'll cover exactly what this means, and why we want this in the next chapter, but first, let's learn about how we can check for it. First, let's use the corr-dot-test() function from psych. This will print a correlation matrix along with significance values for our item correlations. I've printed the correlations of our first three items here. Next, let's visualize the entire matrix with the corrplot() function from the package of the same name. This takes a correlation matrix as input, which we will create with the cor() function from base R. We will use the circle me