Calculate difference in means — get

This function calculates the difference in means using a bivariate regression, as well the p-value indicating how significant each difference is. The main function doing the calculations lm().

NOTE: This function does not perform an actual Dunnet Test as it does not calculate the quantile of the multivariate t-distribution when determining the confidence intervals and p-values. If you need to perform an actual Dunnett Test use the dunnett() function instead. Please be aware that that function is far slower when there are many comparison groups due to the nature of mvtnorm::qmvt() and high dimensional data.

Usage

get_diffs(
  data,
  x,
  treats,
  group,
  wt,
  ref_level = NULL,
  pval_adj = NULL,
  conf_level = 0.95,
  conf_method = c("wald", "profile"),
  show_means = TRUE,
  show_pct_change = FALSE,
  decimals = 3,
  na.rm = TRUE
)

Arguments

data: A data frame or tibble.
x: A numeric vector that will be used to calculate the means. This can be a string or symbol.
treats: A variable whose values are used to determine if the means are statistically significantly different from each other. Should be a factor or character vector. This can be a string or symbol.
group: <tidy-select> A selection of columns to group the data by in addition to treats. This operates very similarly to .by from dplyr (for more info on that see ?dplyr_by). See examples to see how it operates.
wt: Weights. Add if you have a weighting variable and want to perform Dunnett's test with weighted means.
ref_level: A string that specifies the level of the reference group through which the others will be tested.
pval_adj: Method for adjusting p-values for multiple comparisons. Passed directly to stats::p.adjust. Options include "holm", "hochberg", "hommel", "bonferroni", "BH", "BY", "fdr", "none". Default is NULL (no adjustment).
conf_level: A number between 0 and 1 that signifies the width of the desired confidence interval. Default is 0.95, which corresponds to a 95% confidence interval.
conf_method: Determines whether the confidence intervals are calculated using the profile likelihood or the Wald method. Obviously has two options, "profile" and "wald". Wald is between 3 to 25 times as fast but not as reliable for small sample sizes (n < 50). For larger sample sizes, (n > 100), they will be very similar. The default is Wald.
show_means: Logical. Default is FALSE which does not show the mean values for the levels. If TRUE, will add a column called mean that contains the means.
show_pct_change: Logical. Default is FALSE which does not show the percent change from the reference category to the other categories. If TRUE, will show the percent change.
decimals: Number of decimals each number should be rounded to. Default is 3.
na.rm: Logical. Default is TRUE which removes NAs prior to calculation.

Value

A tibble with one row if no group is provided and data is not of class "grouped_df". If data is of class "grouped_df" or group is provided, it will return one row for each unique observation if one group is provides and one row per unique combination of observations if multiple groups are used.