When modeling time‐varying relationships in a longitudinal study, be sure the data are coded consistently for either growth or decline. For example, if a variable is set equal to one value at the start of the study and in subsequent waves, the variable is "1, 0, 0, 1, 0, 1", and not "1, 1, 1, 0, 0, 0". Therefore, if you want to calculate percent change over time, you should set a variable to "1, 0, 0, 1, 0, 1" if you want growth (some intermediate value, say 1, for example), and to "0, 1, 1, 0, 0, 1" for decline (all zeros). Otherwise, doing the same calculations will present you with an incorrect answer, which is that you are seeing decline when in fact you should be seeing growth or vice versa.
If coding a variable as a time‐varying covariate, make sure that you adjust for one or more artificial time trends when formally analyzing time‐varying effects. For example, if your dependent variable is continuous, you may want to set your time covariate to be either linear in time or a quadratic in time (i.e., it is a function of time squared). The other option is to include an intercept in the model.
Including discrete or zero values in calculations should be avoided. If you want to use the calculation "total(price$flat)==7," you need to change the value of flat to be stored as a decimal amount of money. In relation to this, be careful when entering categorical variables as an absolute value, as REDCap treats the field as a radio button, so that all the values in the variable become either 1 or 0. To enter a categorical value as a decimal, you must include the % sign in the characters within the variable name, e.g. "price%".
How do you convert a categorical variable from a patron‐reported field into one that you want to count? Thanks to the use of codebooks, it is very easy to create codes to match your patron‐reported questions into the codes you want to use. d2c66b5586