The following example depicts how to use the scoresheet package for manipulating data and saving some scored data. This example will use 7 items used to assess binge eating behavior at age 14 in the Avon Longitudinal Study of Parents and Children (ALSPAC). You can use load this synthetic data and scoresheet in the package to test the example and familiarize yourself with the package. Building a scoresheet should be an approachable and reproducible process, even for R beginners. While all of this can be accomplished (in some cases more efficiently) with some savvy dplyr coding, the scoresheet format should harmonize the process, increase approachabiity by reducing the need for heavy coding, allow for straightforward and reproducible implementation of scoring practices across studies, and offer a useful way to document and disseminate assessment scoring for publication. This package assumes some basic r knowledge (i.e. how to start up r, load packages and data, assign variables, and use basic computations). Some functions require a minimum of thoughtful coding, and I try to make that process a bit less painful via this vignette. I recommend reading the dplyr documentation as well as the scorekeeper documentation for a specific operation if you are stuck on a particular piece of code.
Ok! Let’s get started. First, we load the scorekeeper package:
Next, let’s familiarize ourselves with the raw data. This example data file includes an id variable, 7 binge eating variables (ccq360, ccq370, ccq371, ccq372, ccq373, ccq374, and ccq375). I have also included 4 ‘noisy’ variables which are related to other assessments of eating behavior, ccq380, ccq381, ccq382 and ccq383.
Raw Data
| id | ccq360 | ccq370 | ccq371 | ccq372 | ccq373 | ccq374 | ccq375 | ccq380 | ccq381 | ccq382 | ccq383 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | 1 | -10 | -10 | -10 | -10 | -10 | -10 | -1 | -1 | -1 | -1 |
| 2 | 2 | 3 | 3 | 3 | 3 | 3 | 3 | 1 | 1 | 1 | 1 |
| 3 | 2 | 1 | 2 | 2 | 1 | 1 | 1 | 1 | 1 | 2 | 1 |
| 4 | -1 | -1 | -1 | -1 | -1 | -1 | -1 | -1 | -1 | -1 | -1 |
| 5 | 2 | 3 | 2 | 1 | 2 | 3 | 1 | 4 | 1 | 1 | 1 |
| 6 | 1 | -10 | -10 | -10 | -10 | -10 | -10 | 1 | 1 | 1 | 1 |
| 7 | 2 | 2 | 2 | 2 | 2 | 3 | 2 | 4 | 2 | 1 | 2 |
| 8 | -1 | -1 | -1 | -1 | -1 | -1 | -1 | -1 | -1 | -1 | -1 |
| 9 | 4 | 3 | 2 | 2 | 3 | 3 | 2 | 1 | 2 | 1 | 4 |
| 10 | 2 | 1 | 3 | 3 | 2 | 3 | 2 | 1 | 1 | 1 | 1 |
As you can see from the examples below, these variables are coded in ways that are slightly odd. For instance, ccq360 is not a descriptive name, and a response of ‘never’ corresponds to the value of 1. There are also two missing data values currently coded as ‘no response’ that I’d like to recode as NA for analysis.
#> B12/B11: Frequency respondent went on an eating binge during the past year (x) <numeric>
#> # total N=10 valid N=10 mean=1.40 sd=1.51
#>
#> Value | Label | N | Raw % | Valid % | Cum. %
#> --------------------------------------------------------------
#> -9999 | Consent withdrawn by YP | 0 | 0 | 0 | 0
#> -11 | Triplet / quadruplet | 0 | 0 | 0 | 0
#> -10 | Not completed | 0 | 0 | 0 | 0
#> -1 | No response | 2 | 20 | 20 | 20
#> 1 | Never | 2 | 20 | 20 | 40
#> 2 | < Once a month | 5 | 50 | 50 | 90
#> 3 | 1-3 times a month | 0 | 0 | 0 | 90
#> 4 | Once a week | 1 | 10 | 10 | 100
#> 5 | <Once a week | 0 | 0 | 0 | 100
#> <NA> | <NA> | 0 | 0 | <NA> | <NA>Similarly, you can see that ccq370 has a ‘no’ answer coded as a 3, with ‘yes, usually’ coded as 1 and ‘yes, sometimes’ coded as a 2. I also see a different missing code pop up here, which is ‘not completed’. If I look at the data patterns and the documentation, I notice that variables ccq370-ccq375 were actually skipped if someone answered that they had never binge eaten in question ccq360. I will also need to recode those variables as ‘No’ instead of ‘not completed’ for those variables in order to account for/fill in what the skip logic in the questionnaire missed
#> B13a/B12a: Respondent felt out of control when on a binge (x) <numeric>
#> # total N=10 valid N=10 mean=-0.90 sd=5.02
#>
#> Value | Label | N | Raw % | Valid % | Cum. %
#> --------------------------------------------------------------
#> -9999 | Consent withdrawn by YP | 0 | 0 | 0 | 0
#> -11 | Triplet / quadruplet | 0 | 0 | 0 | 0
#> -10 | Not completed | 2 | 20 | 20 | 20
#> -1 | No response | 2 | 20 | 20 | 40
#> 1 | Yes, usually | 2 | 20 | 20 | 60
#> 2 | Yes, sometimes | 1 | 10 | 10 | 70
#> 3 | No | 3 | 30 | 30 | 100
#> <NA> | <NA> | 0 | 0 | <NA> | <NA>Scoresheet
When preparing to make a scoresheet - I split my thought process into three chunks (1) cleaning raw variables, (2) creating composite scores, and (3) selecting your final, cleaned variables to save. We’ll start with the first chunk below:
I have identified 4 things that I would like to do to clean raw variables prior to analysis:
1. Select only id, ccq360, ccq370, ccq371, ccq372, ccq373, ccq374, and ccq375 as the variables that I am interested in scoring.
2. Appropriately account for skip patterns and change entries to ‘no’ for variables where a skip pattern was employed.
3. Recode all negative numbers as ‘missing’; recode other responses to be more sensible (e.g. for ccq370: 0 = No, 1 = Yes, sometimes, 2 = Yes, usually for an ordinal variable AND/OR 0 = No and 1 = Yes - sometimes or usually for a dichotomized variable).
4. Rename variables with more descriptive names. Clean up variable labels
Thinking about the order that these will need to be completed: I can do the select operation at any point, but it makes sense to do it first in order to reduce clutter in my data as I manipulate it. I can complete the unskipping (if_else) and recode in either order, but the VALUE that is entered for skipped variables would be different if I completed this before vs. after the recode, as a ‘No’ answer would have a different value before vs. after recoding. It seems most sensible to me to start with filling out the skipped variables, so I will do that step first before recoding. Finally, I would like to rename the variables to be more descriptive names and clean up the variable labels a bit. This renaming and cleaning up labels can be accomplished within the recode step, so I have 3 steps to complete in order to clean my raw variables.
At present, scoresheets should include up to 11 columns which must have names that exactly match the following: raw_vars, new_var, label, operation, step, val_labs, new_vals, if_condition, if_true_return, else_return, code. Not all operations use all columns, so if there is an operation you do not need, you may not need all columns in your scoresheet. Your scoresheet will run regardless of whether your include unnecessary columns, but it will give an error if you do not include columns that are necessary.
Steps 0-3 (below) represent this first chunk of data cleaning - preparing the raw variables:
Step 0
I’ll just document my raw variables of interest in Step 0, so that anyone who is interested in replicating my scoring can verify that their variables are coded as mine are. This will also give me a starting point to refer back to if I would like to remember how the original varialbes were coded before transformation. For example, see the first three rows of raw variables below:
| raw_vars | new_var | label | operation | step | val_labs | new_vals | if_condition | if_true_return | else_return | code |
|---|---|---|---|---|---|---|---|---|---|---|
| ccq360 | NA | B12/B11: Frequency respondent went on an eating binge during the past year | raw | 0 | consent withrdrawn by YP = -9999, triplet/quadruplet = -11, Not completed = -10, No response = -1, Never = 1, < once a month = 2, 1-3 times a month = 3, once a week = 4, > once a week = 5 | NA | NA | NA | NA | NA |
| ccq370 | NA | B13a/B12a: Respondent felt out of control when on a binge (ccq370) | raw | 0 | consent withrdrawn by YP = -9999, triplet/quadruplet = -11, Not completed = -10, No response = -1, Yes, usually = 1, Yes, sometimes = 2, No =3 | NA | NA | NA | NA | NA |
| ccq371 | NA | B13b/B12b: Respondent ate very fast or faster than normal when on a binge | raw | 0 | consent withrdrawn by YP = -9999, triplet/quadruplet = -11, Not completed = -10, No response = -1, Yes, usually = 1, Yes, sometimes = 2, No =3 | NA | NA | NA | NA | NA |
Step 1
I’m going to enter a row in a scoresheet (I use a .csv file in generic spreadsheet software for easy entry and revision, then load it into r using load_csv, make more minor edits, then save your working/final scoresheet as an RData file and/or export as .csv).
To select only the variables that I want, I add a row to my scoresheet with operation = select and step = 1. In the ‘raw_vars’ column, I list the variables that I would like to select in a comma separated charachter string. Any time there is a list of variables for input, they can be in one of two formats - a single, comma separated character string, or a vector of character strings.
note - currently the select operation only works with each element in a scoresheet fully listed. It is not able to accommodate tidyselect actions
| raw_vars | new_var | label | operation | step | val_labs | new_vals | if_condition | if_true_return | else_return | code |
|---|---|---|---|---|---|---|---|---|---|---|
| id, ccq360, ccq370, ccq371, ccq372, ccq373, ccq374, ccq375 | NA | NA | select | 1 | NA | NA | NA | NA | NA | NA |
Now I’ll use this row to select my variables of interest using the scorekeep function, with my raw data file (binge_raw) and scoresheet that contains Steps 0-1 (score1):
|
Step 2
We now have only the 8 variables that we would like to use. Our next step will be to ‘unskip’ the currently skipped variables. To do this, I’m going to use if_else code (see package documentation for additional details). For each row column ccq370-375, I will create a new variable with ’_unskip’ appended to the end of each variable. For example, if ccq370 == 1, I’ll change the value of the skipped variable to 3; otherwise, I’ll keep the original value. See below for example of this code which creates a new variable ‘ccq370_unskip’:
note - the if_else operation is currently designed to recode one variable per row. Vectorization may be available in future releases
| raw_vars | new_var | label | operation | step | val_labs | new_vals | if_condition | if_true_return | else_return | code |
|---|---|---|---|---|---|---|---|---|---|---|
| ccq370 | ccq370_unskip | NA | if_else | 2 | NA | NA | ccq360 == 1 | 3 | as.numeric(ccq370) | NA |
Now we’ll expand our scoring to Steps 1-2
#complete the `scorekeep` function using my raw data and the first two scoresheet steps:
knitr::kable (scorekeep(binge_raw, score2))
|
|
You’ll see that we now have 2 tibbles in our output, one for each step. The first step selects the variables, and the second step corrects the skip logic.
Note - the scorekeeper package is designed to be very conservative in variable conservation. What this means is that original variables are never discarded, and output tibbles will generally lengthen as you create new variables. There are two operations where this is not the case. First, the select function will reduce columns in a tibble by selecting only those that are identified. Second, the rename function renames a column in a tibble directly. All other operations will generally require a ‘new_var’ name that is NOT yet a named variable and will append the newly created variable to the end of a tibble. You can always select your final variables and rename them back to an original name as a final step in the scorekeeping process if you prefer the original name of a variable but need to make modifications to the data.
Step 3
Now we’re ready for Step 3, recoding and renaming. Below is an example of what a scoresheet row will look like for this operation
| raw_vars | new_var | label | operation | step | val_labs | new_vals | if_condition | if_true_return | else_return | code |
|---|---|---|---|---|---|---|---|---|---|---|
| ccq370_unskip | binge_loc_d.14 | Eating binges felt out of control - present or absent | recode | 3 | no = 0, yes - sometimes or usually = 1 | 1 = 1, 2 = 1, 3 = 0, -1 = NA | NA | NA | NA | NA |
I am using the newly minted ccq370_unskip variable when recoding. I rename the variable something more sensible binge_loc_d.14 - which notes that it’s a binge eating variable assessing loss of control, dichotomized, assessed at age 14. I edited the label just a bit as well. The new value labels are coded as ‘label’ = ‘value’, and the new values recode the old values under the pattern ‘old’ = ‘new’, with commas separating each element in the recoding scheme. Recoded variables will also include a companion variable ‘new_var.factor’, which includes the value label easily visible, so that you can double check that your recoding scheme is correct.
Note - do not use commas within the value labels (e.g. ‘yes, sometimes or usually’ includes a comma and would not parse as an accurate value. Other punctuation may be used within value labels)
Beginning with this step, I’m only going to show the parts of output for readability – as noted above, the scorekeep function scores iteratively and saves each step as a tibble in your output so that you can ensure that the correct transformations were made on each step and troubleshoot if necessary.
| id | ccq360 | ccq370 | ccq371 | ccq372 | ccq373 | ccq374 | ccq375 | ccq370_unskip | ccq371_unskip | ccq372_unskip | ccq373_unskip | ccq374_unskip | ccq375_unskip | binge_freq_d.14 | binge_freq_d.14.factor | binge_loc_d.14 | binge_loc_d.14.factor | binge_fast_d.14 | binge_fast_d.14.factor | binge_stomach_d.14 | binge_stomach_d.14.factor | binge_large_d.14 | binge_large_d.14.factor | binge_hide_d.14 | binge_hide_d.14.factor | binge_guilt_d.14 | binge_guilt_d.14.factor |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | 1 | -10 | -10 | -10 | -10 | -10 | -10 | 3 | 3 | 3 | 3 | 3 | 3 | 0 | <1x/mo; absent | 0 | no | 0 | no | 0 | no | 0 | no | 0 | no | 0 | no |
| 2 | 2 | 3 | 3 | 3 | 3 | 3 | 3 | 3 | 3 | 3 | 3 | 3 | 3 | 0 | <1x/mo; absent | 0 | no | 0 | no | 0 | no | 0 | no | 0 | no | 0 | no |
| 3 | 2 | 1 | 2 | 2 | 1 | 1 | 1 | 1 | 2 | 2 | 1 | 1 | 1 | 0 | <1x/mo; absent | 1 | yes - sometimes or usually | 1 | yes - sometimes or usually | 1 | yes - sometimes or usually | 1 | yes - sometimes or usually | 1 | yes - sometimes or usually | 1 | yes - sometimes or usually |
| 4 | -1 | -1 | -1 | -1 | -1 | -1 | -1 | -1 | -1 | -1 | -1 | -1 | -1 | NA | NA | NA | NA | NA | NA | NA | NA | NA | NA | NA | NA | NA | NA |
| 5 | 2 | 3 | 2 | 1 | 2 | 3 | 1 | 3 | 2 | 1 | 2 | 3 | 1 | 0 | <1x/mo; absent | 0 | no | 1 | yes - sometimes or usually | 1 | yes - sometimes or usually | 1 | yes - sometimes or usually | 0 | no | 1 | yes - sometimes or usually |
| 6 | 1 | -10 | -10 | -10 | -10 | -10 | -10 | 3 | 3 | 3 | 3 | 3 | 3 | 0 | <1x/mo; absent | 0 | no | 0 | no | 0 | no | 0 | no | 0 | no | 0 | no |
| 7 | 2 | 2 | 2 | 2 | 2 | 3 | 2 | 2 | 2 | 2 | 2 | 3 | 2 | 0 | <1x/mo; absent | 1 | yes - sometimes or usually | 1 | yes - sometimes or usually | 1 | yes - sometimes or usually | 1 | yes - sometimes or usually | 0 | no | 1 | yes - sometimes or usually |
| 8 | -1 | -1 | -1 | -1 | -1 | -1 | -1 | -1 | -1 | -1 | -1 | -1 | -1 | NA | NA | NA | NA | NA | NA | NA | NA | NA | NA | NA | NA | NA | NA |
| 9 | 4 | 3 | 2 | 2 | 3 | 3 | 2 | 3 | 2 | 2 | 3 | 3 | 2 | 1 | 1x/mo or more; present | 0 | no | 1 | yes - sometimes or usually | 1 | yes - sometimes or usually | 0 | no | 0 | no | 1 | yes - sometimes or usually |
| 10 | 2 | 1 | 3 | 3 | 2 | 3 | 2 | 1 | 3 | 3 | 2 | 3 | 2 | 0 | <1x/mo; absent | 1 | yes - sometimes or usually | 0 | no | 0 | no | 1 | yes - sometimes or usually | 0 | no | 1 | yes - sometimes or usually |
This concludes the first chunk of data cleaning - recoding the raw variables – in this example. Our variables binge_freq_d.14 through the end of the tibble now represent appropriately coded (dichotomized) variables for our analysis. For the second chunk of data cleaning, we’ll create composite scores. The composite scores that I have decided to create are a sum of symptoms that are present: binge_sx_sum.14, a variable that dichotomizes binge eating issues as present if individuals report binge at least once per month with at least two of the six additional symptoms, and a variable that splits the sum score into categories of absent, mild, and severe. We’ll need to first create the sum score before we can complete the next two operations.
Step 4
Our next step will be to create a sum score. Below is the scoresheet row to create our symptom sum score, which gives a count of the number of binge eating symptoms (0-6) that are present for each individual
| raw_vars | new_var | label | operation | step | val_labs | new_vals | if_condition | if_true_return | else_return | code |
|---|---|---|---|---|---|---|---|---|---|---|
| binge_loc_d.14, binge_fast_d.14, binge_stomach_d.14, binge_large_d.14, binge_hide_d.14, binge_guilt_d.14 | binge_sx_sum.14 | Sum of binge eating symtpoms Age 14 | sum | 4 | NA | NA | NA | NA | NA | NA |
The sum operation computes five variables, the raw sum of the list of variables provided in raw_vars (NAs excluded), the raw sum of those with complete data only, the number of missing values, the percent of values that are missing, and a weighted sum (weighted by the number of non-missing values). Below are the last 5 columns of the tibble created in Step 4:
| binge_sx_sum.14 | binge_sx_sum.14_complete | binge_sx_sum.14_NAs | binge_sx_sum.14_NA_percent | binge_sx_sum.14_weighted_sum |
|---|---|---|---|---|
| 0 | 0 | 0 | 0 | 0 |
| 0 | 0 | 0 | 0 | 0 |
| 6 | 6 | 0 | 0 | 6 |
| 0 | NA | 6 | 100 | NA |
| 4 | 4 | 0 | 0 | 4 |
| 0 | 0 | 0 | 0 | 0 |
| 5 | 5 | 0 | 0 | 5 |
| 0 | NA | 6 | 100 | NA |
| 3 | 3 | 0 | 0 | 3 |
| 3 | 3 | 0 | 0 | 3 |
Step 5
In step 5, we create two additional variables based on the sum score created in step 4, and we filter the tibble to exclude cases with all missing data. We can do all of these operations in the same step because they are independent and do not rely on one another. The if_else operation computes a variable that notes binge eating is present (‘1’) if an individual reports binge eating at least once per month with more than two additional symptoms. The case_when operation splits the sum score into ‘absent’, ‘mild’, and ‘severe’ cases based on the number of symtpoms reported. See if_else, filter_at and case_when documentation in the package help files for more information.
| raw_vars | new_var | label | operation | step | val_labs | new_vals | if_condition | if_true_return | else_return | code |
|---|---|---|---|---|---|---|---|---|---|---|
| binge_sx_sum.14, binge_freq_d.14 | binge_present.14 | binge eating present and more than 2 symptoms | if_else | 5 | present = 1, absent = 0 | NA | binge_freq_d.14 == 1 & binge_sx_sum.14 > 2 | 1 | 0 | NA |
| binge_sx_sum.14 | binge_sx_sum_split.14 | Binge symptoms sum split into absent, mild, severe | case_when | 5 | absent = 0, mild = 1, severe = 2 | NA | NA | NA | NA | binge_sx_sum.14 == 0 ~ 0, binge_sx_sum.14 > 0 & binge_sx_sum.14 <=3 ~ 1, binge_sx_sum.14 > 3 ~ 2 |
| NA | NA | NA | filter_at | 5 | NA | NA | NA | NA | NA | c(15:28), any_vars(!is.na(.)) |
The last two variables created are seen below:
| binge_present.14 | binge_sx_sum_split.14 |
|---|---|
| 0 | 0 |
| 0 | 0 |
| 0 | 2 |
| 0 | 0 |
| 0 | 2 |
| 0 | 0 |
| 0 | 2 |
| 0 | 0 |
| 1 | 1 |
| 0 | 1 |
Steps 6-7
Now we’ve created the final composite scores. In our final chunk, we’ll clean up the dataset. In step 6, we select only a few columns that will go into the final dataset, and in step 7, we rename binge_sx_sum_split.14 to a simpler name: binge_severity.14
| raw_vars | new_var | label | operation | step | val_labs | new_vals | if_condition | if_true_return | else_return | code |
|---|---|---|---|---|---|---|---|---|---|---|
| id, binge_freq_d.14, binge_loc_d.14, binge_fast_d.14, binge_stomach_d.14, binge_large_d.14, binge_hide_d.14, binge_guilt_d.14, binge_present.14, binge_sx_sum.14, binge_sx_sum_split.14 | NA | NA | select | 6 | NA | NA | NA | NA | NA | NA |
| binge_sx_sum_split.14 | binge_severity.14 | NA | rename | 7 | NA | NA | NA | NA | NA | NA |
The Final Reveal
Putting it all together: we now have a full scoresheet -
| raw_vars | new_var | label | operation | step | val_labs | new_vals | if_condition | if_true_return | else_return | code |
|---|---|---|---|---|---|---|---|---|---|---|
| ccq360 | NA | B12/B11: Frequency respondent went on an eating binge during the past year | raw | 0 | consent withrdrawn by YP = -9999, triplet/quadruplet = -11, Not completed = -10, No response = -1, Never = 1, < once a month = 2, 1-3 times a month = 3, once a week = 4, > once a week = 5 | NA | NA | NA | NA | NA |
| ccq370 | NA | B13a/B12a: Respondent felt out of control when on a binge (ccq370) | raw | 0 | consent withrdrawn by YP = -9999, triplet/quadruplet = -11, Not completed = -10, No response = -1, Yes, usually = 1, Yes, sometimes = 2, No =3 | NA | NA | NA | NA | NA |
| ccq371 | NA | B13b/B12b: Respondent ate very fast or faster than normal when on a binge | raw | 0 | consent withrdrawn by YP = -9999, triplet/quadruplet = -11, Not completed = -10, No response = -1, Yes, usually = 1, Yes, sometimes = 2, No =3 | NA | NA | NA | NA | NA |
| ccq372 | NA | B13c/B12c: Respondent ate until their stomach hurt or they felt sick when on a binge | raw | 0 | consent withrdrawn by YP = -9999, triplet/quadruplet = -11, Not completed = -10, No response = -1, Yes, usually = 1, Yes, sometimes = 2, No =3 | NA | NA | NA | NA | NA |
| ccq373 | NA | B13d/B12d: Respondent ate really large amounts of food when not hungry when on a binge (ccq373) | raw | 0 | consent withrdrawn by YP = -9999, triplet/quadruplet = -11, Not completed = -10, No response = -1, Yes, usually = 1, Yes, sometimes = 2, No =3 | NA | NA | NA | NA | NA |
| ccq374 | NA | B13e/B12e: Respondent ate by themselves to hide amount eaten when on a binge (ccq374) | raw | 0 | consent withrdrawn by YP = -9999, triplet/quadruplet = -11, Not completed = -10, No response = -1, Yes, usually = 1, Yes, sometimes = 2, No =3 | NA | NA | NA | NA | NA |
| ccq375 | NA | B13f/B12f: Respondent felt really bad / guilty after eating a lot of food when on a binge (ccq375) | raw | 0 | consent withrdrawn by YP = -9999, triplet/quadruplet = -11, Not completed = -10, No response = -1, Yes, usually = 1, Yes, sometimes = 2, No =3 | NA | NA | NA | NA | NA |
| id, ccq360, ccq370, ccq371, ccq372, ccq373, ccq374, ccq375 | NA | NA | select | 1 | NA | NA | NA | NA | NA | NA |
| ccq370 | ccq370_unskip | NA | if_else | 2 | NA | NA | ccq360 == 1 | 3 | as.numeric(ccq370) | NA |
| ccq371 | ccq371_unskip | NA | if_else | 2 | NA | NA | ccq360 == 1 | 3 | as.numeric(ccq371) | NA |
| ccq372 | ccq372_unskip | NA | if_else | 2 | NA | NA | ccq360 == 1 | 3 | as.numeric(ccq372) | NA |
| ccq373 | ccq373_unskip | NA | if_else | 2 | NA | NA | ccq360 == 1 | 3 | as.numeric(ccq373) | NA |
| ccq374 | ccq374_unskip | NA | if_else | 2 | NA | NA | ccq360 == 1 | 3 | as.numeric(ccq374) | NA |
| ccq375 | ccq375_unskip | NA | if_else | 2 | NA | NA | ccq360 == 1 | 3 | as.numeric(ccq375) | NA |
| ccq360 | binge_freq_d.14 | Eating binges (eating a large amount of food) present or absent in the past year’ | recode | 3 | <1x/mo; absent = 0, 1x/mo or more; present = 1 | 1 = 0, 2= 0, 3 = 1, 4 = 1, 5 = 1, -1 = NA | NA | NA | NA | NA |
| ccq370_unskip | binge_loc_d.14 | Eating binges felt out of control - present or absent | recode | 3 | no = 0, yes - sometimes or usually = 1 | 1 = 1, 2 = 1, 3 = 0, -1 = NA | NA | NA | NA | NA |
| ccq371_unskip | binge_fast_d.14 | Eating Very fast or faster than normal when on a binge - present or absent | recode | 3 | no = 0, yes - sometimes or usually = 1 | 1 = 1, 2 = 1, 3 = 0, -1 = NA | NA | NA | NA | NA |
| ccq372_unskip | binge_stomach_d.14 | Eating until Stomach hurt or felt sick - present or absent | recode | 3 | no = 0, yes - sometimes or usually = 1 | 1 = 1, 2 = 1, 3 = 0, -1 = NA | NA | NA | NA | NA |
| ccq373_unskip | binge_large_d.14 | Eating large amounts of food when not hungry - present or absent | recode | 3 | no = 0, yes - sometimes or usually = 1 | 1 = 1, 2 = 1, 3 = 0, -1 = NA | NA | NA | NA | NA |
| ccq374_unskip | binge_hide_d.14 | Hiding amount eaten when on a binge - present or absent | recode | 3 | no = 0, yes - sometimes or usually = 1 | 1 = 1, 2 = 1, 3 = 0, -1 = NA | NA | NA | NA | NA |
| ccq375_unskip | binge_guilt_d.14 | Feeling bad or guilty after a binge - present or absent | recode | 3 | no = 0, yes - sometimes or usually = 1 | 1 = 1, 2 = 1, 3 = 0, -1 = NA | NA | NA | NA | NA |
| binge_loc_d.14, binge_fast_d.14, binge_stomach_d.14, binge_large_d.14, binge_hide_d.14, binge_guilt_d.14 | binge_sx_sum.14 | Sum of binge eating symtpoms Age 14 | sum | 4 | NA | NA | NA | NA | NA | NA |
| binge_sx_sum.14, binge_freq_d.14 | binge_present.14 | binge eating present and more than 2 symptoms | if_else | 5 | present = 1, absent = 0 | NA | binge_freq_d.14 == 1 & binge_sx_sum.14 > 2 | 1 | 0 | NA |
| binge_sx_sum.14 | binge_sx_sum_split.14 | Binge symptoms sum split into absent, mild, severe | case_when | 5 | absent = 0, mild = 1, severe = 2 | NA | NA | NA | NA | binge_sx_sum.14 == 0 ~ 0, binge_sx_sum.14 > 0 & binge_sx_sum.14 <=3 ~ 1, binge_sx_sum.14 > 3 ~ 2 |
| NA | NA | NA | filter_at | 5 | NA | NA | NA | NA | NA | c(15:28), any_vars(!is.na(.)) |
| id, binge_freq_d.14, binge_loc_d.14, binge_fast_d.14, binge_stomach_d.14, binge_large_d.14, binge_hide_d.14, binge_guilt_d.14, binge_present.14, binge_sx_sum.14, binge_sx_sum_split.14 | NA | NA | select | 6 | NA | NA | NA | NA | NA | NA |
| binge_sx_sum_split.14 | binge_severity.14 | NA | rename | 7 | NA | NA | NA | NA | NA | NA |
And we can use that scoresheet to fully score our binge eating measure with just one line of code:
cleaned_data <- scorekeep(binge_raw, binge_scoresheet)The last tibble in this set of 7 tibbles will be the fully cleaned data
| id | binge_freq_d.14 | binge_loc_d.14 | binge_fast_d.14 | binge_stomach_d.14 | binge_large_d.14 | binge_hide_d.14 | binge_guilt_d.14 | binge_present.14 | binge_sx_sum.14 | binge_severity.14 |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| 2 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| 3 | 0 | 1 | 1 | 1 | 1 | 1 | 1 | 0 | 6 | 2 |
| 5 | 0 | 0 | 1 | 1 | 1 | 0 | 1 | 0 | 4 | 2 |
| 6 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| 7 | 0 | 1 | 1 | 1 | 1 | 0 | 1 | 0 | 5 | 2 |
| 9 | 1 | 0 | 1 | 1 | 0 | 0 | 1 | 1 | 3 | 1 |
| 10 | 0 | 1 | 0 | 0 | 1 | 0 | 1 | 0 | 3 | 1 |
And we can start to analyze our cleaned variables, which I can then save, export, and connect with other data files :)
#> Binge symptoms sum split into absent, mild, severe (x) <numeric>
#> # total N=8 valid N=8 mean=1.00 sd=0.93
#>
#> Value | Label | N | Raw % | Valid % | Cum. %
#> ---------------------------------------------
#> 0 | absent | 3 | 37.50 | 37.50 | 37.50
#> 1 | mild | 2 | 25.00 | 25.00 | 62.50
#> 2 | severe | 3 | 37.50 | 37.50 | 100.00
#> <NA> | <NA> | 0 | 0.00 | <NA> | <NA>While developing this initial scoresheet was somewhat labor intensive, the scoresheet creates a documented, reproducible scoring and data cleaning system that can be easily shared. For instance, if another individual has data on the same measure, they can use my scoresheet (which I can save to a .csv file and post/send) to clean and score their data in the exact same way that I did, using one line of code. If raw variables are named slightly differently across projects, I would only need to adjust the raw variable names in a few places in the scoresheet for it to work on new data. While the scorekeeper package can be used for any multi-item scoring tasks, it shines for algorithmic tasks (i.e. deriving psychological diagnoses), which often involve multiple if_else or case_when statements.
Happy scoring!