aMNLFA.prune() shows which effects will be significant based on different thresholds.
First, it will give you alpha levels corresponding to two possible options for how to the number of tests m: (1) set as the number of all possible tests (i.e., the number of items multiplied by the number of covariates; shown as “ibc” in the results); or (2) the number of tests actually included in the simultaneous model, shown as “actual” in the results.
Second, it will give you the option to use either (1) a Benjamini-Hochberg correction, shown as “BH” in the results; or (2) a Bonferroni correction, shown as “bon” in the results.
Third, if you’re testing threshold DIF, it will give you the results if you trim non-significant effects based on (1) all categories’ thresholds for each item, shown as “tdif” in the results; and (2) the category with the highest test statistic for each item, shown as “intdif” in the results.
The main difference here is that the aMNLFA.prune() function creates a different type of object (i.e., not an aMNLFA.object), which can be inspected and passed to aMNLFA.plot(). This function will give you plots for each different type of DIF parameter you have, as well as how your decisions would change based on different thresholds.
The new order of operations is as follows:
- Set up an aMNLFA.object.
- Run aMNLFA.sample().
- Run aMNLFA.initial() and the corresponding .inp file in Mplus.
- Run aMNLFA.simultaneous() and the corresponding .inp file in Mplus.
- Run aMNLFA.prune() and feed the results of aMNLFA.prune() to aMNLFA.DIFplot() to visualize results. Decide on the method of trimming effects for the final model.
Example:
prune.object <- aMNLFA.prune(some.object)
aMNLFA.DIFplot(prune.object, “loading”, log = FALSE)
aMNLFA.DIFplot(prune.object, “intercept”, log = FALSE)
- Run aMNLFA.final(), adding arguments for the adjustment method (method = c(“bh”, “bonferroni”)), the way of choosing the number of tests (mchoice = c(“actual”, “ibc”)), and, if you’re using thresholds, which categories you want to test (highest.category = c(TRUE, FALSE)). Run the corresponding .inp file in Mplus.
Example:
aMNLFA.final(some.object, mchoice = “actual”, method = “bh”)