Description Usage Arguments Details Value Author(s) See Also Examples
Estimate summary statistics and effect size comparisons for each independent set of replicates in a structured experiment. Choice of functions for summary statistics and comparisons are flexible. Inputs use a formula interface and human-understandable variable types such as responses, treatments, time, and block.
1 2 3 4 5 6 | treateffect(data, formula = NULL,
response = NULL, treatment = NULL, groups = NULL, times = NULL, block = NULL,
pool_variance = NULL, average_subsamples = FALSE,
summary_functions = c("mean", "se", "CI68"), comp_groups = allpairwise,
control = NULL, comp_function = NULL, conf.int = 0.95,
CI_derivation = "ML", effect_size_type = "difference")
|
data |
a data frame |
formula |
The formula is specified as Using a double underscore, treatments or groups can be tagged for variance pooling as |
response |
If a formula is not specified, the response variables y1 + y2 + ... can be specified as a character vector. |
treatment |
If a formula is not specified, the treatment variables x1 + x2 + ... can be specified as a character vector. |
groups |
If a formula is not specified, the response variables g1 + g2 + ... can be specified as a character vector. |
times |
an optional character string indicating the name of a time variable (which can be numeric or of any of the typical date and time formats). Only one time variable can be specified. |
block |
an optional character string indicating the names of variables indicating blocking or pairing structure in the data. This is used by |
pool_variance |
An optional character vector indicating over which variables to pool the variance. This information can be used by comp_function functions that pool variance such as |
average_subsamples |
a logical argument indicating whether any remaining replicates should be averaged. |
summary_functions |
a character vector including the name of one or multiple functions to be used to summarize each treatment at each time point within each group. Functions |
comp_groups |
a function that will be used to set up specific comparisons to be made. The three currently available functions are |
control |
an optional character vector the length of the number of specified treatement variables) indicating which treatment level(s) is/are the "control" for each treatment variable if comparisons with the control are desired. If nothing is specified, the first level in each treatment vector is used. Create an ordered variable or specify a control to change which level will be used as the control. |
comp_function |
a function that will be used to make comparisons. See |
conf.int |
The alpha level for confidence interval calculations. CI_derivation = "ML", effect_size_type = "difference" |
CI_derivation |
character vector that will be used by |
effect_size_type |
chacter vector specifying how an effect size be reported. The two options are: "difference" (the default) and "ratio" Keep in mind that ratios often do not make sense for data that contain 0 or negative values. |
Using a formula interface, treateffect facilitates efficient calculation of the size of treatment effect sizes when comparing multiple categorical treatments. Standard variable types such as response variables (of which there can be multiple for efficient analysis), treatment categories (multiple variables also allowed), a time variable, a blocking variable, variables over which to pool the data, and panel variables over which to divide the data before analysis (e.g. two sites) can be specified. Functions used to calculate the size of the effect are flexible. The specific summary statistics (e.g., mean, SE, SD), comparison function (e.g., confidence interval from Welch t-test or a bootstrapped confidence interval) and the comparisons to perform (e.g., all pairwise comparisons or multiple comparisons with a control) can be specified.
Returns an object of class "te
". A print
method shows the results in tabular form and the plot
and plotdiff
, can be used to plot the results.
An object of class "te
" is a list containing some or all of the following components:
source_data |
the unmodified data supplied via the |
design |
a list of all of the "design" features used to shape the analysis and plotting such as the identities of the treatment, response, time, and block variables |
data |
the data frame used for analysis after restructuring to accommodate for example multiple response variables |
treatment_summaries |
one of two main output data frames showing the output of the |
treatment_comparisons |
the second of two main output data frames showing the output of the analyses by the specified |
Anthony Darrouzet-Nardi
plot.te
, plotdiff
, define_comparisons
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 | theme_te() #a more spartan aesthetic
### EXAMPLES WITH SIMULATED DATA
# very basic case: 1 response, 1 treatment with 2 levels
ex1 <- tedatasim(n = 10, response = 1, levels = 2, time = 1,
groups = 1, subsample = 1, block = FALSE)
ex1.te <- treateffect(ex1, resp_var1 ~ pred_var1)
ex1.te
plot(ex1.te)
plotdiff(ex1.te)
# 2 responses
ex2 <- tedatasim(n = 10, response = 2, levels = 2, time = 1,
groups = 1, subsample = 1, block = FALSE)
ex2.te <- treateffect(ex2, resp_var1 + resp_var2 ~ pred_var1)
ex2.te
plot(ex2.te)
plotdiff(ex2.te)
# 2 predictors
ex3 <- tedatasim(n = 3, response = 1, levels = c(2,2), time = 1,
groups = 1, subsample = 1, block = FALSE)
ex3.te <- treateffect(ex3, resp_var1 ~ pred_var1 + pred_var2)
ex3.te
plot(ex3.te)
plotdiff(ex3.te)
# univariate data
ex4.te <- tedatasim(n = 10, response = 1, levels = 1, time = 1,
groups = 1, subsample = 1, block = FALSE) %>%
treateffect(resp_var1 ~ pred_var1)
ex4.te
plot(ex4.te)
## Examples with time variables
# One treatment over time
ex5 <- tedatasim(n = 5, response = 1, levels = 1, time = 3,
groups = 1, subsample = 1, block = FALSE)
ex5.te <- treateffect(ex5, resp_var1 ~ pred_var1 | time_var__time)
ex5.te
plot(ex5.te)
# Two treatments over time
ex6 <- tedatasim(n = 5, response = 1, levels = 2, time = 10,
groups = 1, subsample = 1, block = FALSE)
ex6.te <- treateffect(ex6, resp_var1 ~ pred_var1 | time_var__time)
ex6.te
plot(ex6.te, dodge = 0.3)
plotdiff(ex6.te, dodge = 0.3)
# 3 groups, + time
ex7 <- tedatasim(n = 10, response = 1, levels = 2, time = 30,
groups = 3, subsample = 1, block = FALSE)
ex7.te <- treateffect(ex7, resp_var1 ~ pred_var1 | group_var1 + time_var__time)
ex7.te
plot(ex7.te, panel_formula = group_var1 ~ .)
## EXAMPLES WITH VARIOUS INTERNAL DATA SETS
#warpbreaks - a commonly used R data set
wb.te <- treateffect(warpbreaks, breaks ~ wool:tension)
wb.te
plot(wb.te)
plotdiff(wb.te)
#barley (using comparisons with all other treatments)
barley.te <- lattice::barley %>%
treateffect(yield ~ variety__pool | year, comp_groups = allothers)
barley.te
plot(barley.te)
plotdiff(barley.te)
#city example used in many bootstrapping examples
city_tidy <- boot::city %>%
gather(year, population) %>%
mutate(year = factor(year, labels = c(1920, 1930)), city = rep(1:10, 2))
treateffect(city_tidy, population ~ year + city__block,
comp_function = bootRR_bca_paired)
#pool variance among treatments with Tukey test
amod <- aov(breaks ~ tension, data = warpbreaks)
confint(multcomp::glht(amod, linfct = mcp(tension = "Tukey")))$confint
treateffect(warpbreaks, breaks ~ tension__pool)
#Dunnett multiple comparisons with control example
amod <- aov(breaks ~ tension, data = warpbreaks)
confint(multcomp::glht(amod, linfct = mcp(tension = "Dunnett")))$confint
treateffect(warpbreaks, breaks ~ tension__pool, comp_groups = mcc)
#specify a different control
treateffect(warpbreaks, breaks ~ tension__pool, comp_groups = mcc,
control = "H")
#Bayesian BESTmcmc "supersedes the t-test"
y1 <- c(5.77, 5.33, 4.59, 4.33, 3.66, 4.48)
y2 <- c(3.88, 3.55, 3.29, 2.59, 2.33, 3.59)
summary(BESTmcmc(y1,y2))
data.frame(r = c(y1,y2), t = rep(c("y1","y2"), ea = 6)) %>%
treateffect(r ~ t, comp_function = BESTHDI)
#starwars dataset from tidyr with lots missing and categories with 1 number
treateffect(starwars, height + mass + birth_year ~ gender) %>% plot
#Lots of ways to skin a cat
treateffect(NSE_7mo, NSE_7mo ~ treatment | year) %>% plot
treateffect(NSE_7mo, NSE_7mo ~ treatment) %>% plotdiff
treateffect(NSE_7mo, NSE_7mo ~ treatment | year) %>% plotdiff
treateffect(NSE_7mo, NSE_7mo ~ treatment | year + block__block) %>% plotdiff
treateffect(NSE_7mo, NSE_7mo ~ treatment__pool | year + block__block) %>% plotdiff
treateffect(NSE_7mo, NSE_7mo ~ treatment__pool | year__pool +
block__block) %>% plotdiff
treateffect(NSE_7mo, NSE_7mo ~ treatment__pool | year__pool) %>% plotdiff
treateffect(NSE_7mo, NSE_7mo ~ treatment | year,
comp_function = bootdiff_bca) %>% plotdiff
treateffect(NSE_7mo, NSE_7mo ~ treatment | year__time + block__block,
comp_groups = mcc) %>% plotdiff(dodge = 0.5)
#this next one gives wrong answers and I should find out why bc someone will try it.
treateffect(NSE_7mo, NSE_7mo ~ treatment | year__pool + block__block) %>% plotdiff
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.