Big Five Example - BiFiX

In this vignette, you will get a complete Walkthrough of how to use rextor for your own data analysis. We will use Open Source data on Big Five personality traits, available from the iScience Konstanz account on GitHub: https://github.com/iscience-kn/BiFiX.

Step 1: Reading in the Data

We will be reading in the data directly from the iScience GitHub Account. Make sure to select the raw data file on GitHub or just copy the code below.

In this dataset, we do not need the validation variable. It facilitates dropout analysis in some cases (especially in one-item-one-screen designs) but here we will focus on other variables to clean the data¹. In the first step, we will read in the WEXTOR based data and assign in to a variable named “raw”, that will become our data frame.

raw <- read_WEXTOR(path_to_file("BiFiX_data_raw.csv"),
                    keep_validation = FALSE)

Step 2: Data Cleaning

Great! We now have our empirical data in R, ready to be worked with. But there are some things which are not yet ideal, so will use some more rextor functions to make our lives easier.

When you open up the data and look at it, you might notice that the first variables in the data, which are automatically recorded server variables have the prefix “.wx.”. In R, we generally want our variable names to start with a letter for easy access, so we can either use the namepref function (for variable name prefix) to change that to “v_”. Notice that this function will not add that prefix to the variables that were deliberately named by someone in WEXTOR. Another option is to remove server-generated variable prefixes altogether, which is the option we will choose here. It will only change those variable names that actually have the old prefix that you specify as shown.

data <- removepref(raw) # namepref(raw, ".wx.", "v_")

Seriousness check

We will also perform a “seriousness check” and filter out data of anyone who did not confirm they want to participate seriously. This has proven to improve data quality.

CAVE: With this step you will remove likely remove a lot of your data so make sure you always have a safety copy of the original data. I recommend keeping the raw dataset in your R environment while you prepare your data so you can go back and repeat or adjust any step whenever necessary. That way you can also look at every row of data in your raw data that was removed in later steps and manually check if that filtering was correct or if you need to add some of the data back in.

data2 <- serious_check(data)

Plausibility check

Similar to the seriousness check, plausibility checks are essential in online studies to ensure data quality because researchers have less control over participants’ environments and behavior than in lab settings. Responses may include inattentive participants, bots, or repeated submissions, all of which can bias results if left unchecked. Two particularly effective starting points are session length and IP-based duplicate checks. Extremely short or long session durations can indicate careless responding or technical issues, while repeated IP addresses may signal multiple submissions from the same source. Together, these checks provide a quick, low-threshold way to flag clearly problematic cases before applying more fine-grained data screening procedures. With rextor, the checks described above can be performed separately (as shown below in the commented out part) or in one meta-function. Possibly implausible cases will be flagged as such so the researcher can make an informed decision on whether to keep these cases or exclude them for better data quality. As the BiFiX dataset does not include real IP adresses for data security reasons, we will add simulated ones to show the functionality of the IP check:

# Add random numbers to act as simulated IP addresses
data2$ip <- sample(1:1000, nrow(data2), replace = TRUE)

# Perform separate checks
# data3 <- sess_length_check(data2, 
#                      min_pages = 6)
# data3 <- ip_check(data3)

# Perform overall plausibility check
data3 <- plausicheck(data2, 
                     min_pages = 6,
                     check_sess_length = TRUE,
                     check_ip = TRUE)

Pipe it all together

When preparing data for further analyses it makes sense to have an unchanged raw version of the data in R and then save any changes in a new data frame. However, in order to avoid situations like above with data1, data2 etc., you might want to use pipes to perform several data preparation steps in one go. Since version 4.1.0, R offers a native pipe that looks like this: |>. It allows you to take the output from previous functions as the first input of the next function and thus create a workflow pipeline that is easily adjustable and usually very efficient and clear to read. To avoid having to add simulated IPs here again, we will skip this check but still perform the meta function plausicheck to get the overall plausibilty variable added to the data.

final_data <- raw |> 
  removepref() |> 
  serious_check() |> 
  plausicheck(min_pages = 6, check_sess_length = TRUE, check_ip = FALSE)