its_lm_fourier: ITS analysis for continuous outcomes with seasonal...

In Yael-Travis-Lumer/its2es: Interuppted Time Series Analysis and Corresponding Effect Sizes

ITS analysis for continuous outcomes with seasonal adjustments via Fourier terms.

Description

its_lm_fourier fits a linear regression model asjusted to seasonality, and returns the model, the summary of the model (including the mean difference and Cohen's d), and the original data together with the model predictions.

Usage

its_lm_fourier(
  data,
  form,
  time_name,
  intervention_start_ind,
  freq,
  keep_significant_fourier = TRUE,
  impact_model = "full",
  counterfactual = FALSE,
  print_summary = FALSE
)

Arguments

data	The data frame corresponding to the supplied formula, existing of at least 2 variables: (1) the count outcome, and (2) a vector of time points
form	A formula with the response on the left, followed by the ~ operator, and the covariates on the right, separated by + operators. The formula should not contain an offest term.
time_name	A string giving the name of the time variable. The time variable may or may not be supplied as a covariate in the formula
intervention_start_ind	Numeric - a number between 1 and nrow(data)-1 stating the time point of the start of the intervention
freq	A positive integer describing the frequency of the time series.
keep_significant_fourier	Logical - indicating whether only the significant Fourier terms should be considered. Default is TRUE and then the model is fitted twice; once to obtain the significant Fourier terms, and second time keeping only the significant Fourier terms. If FALSE, then all the Fourier terms are used.
impact_model	A string specifying the assumed impact model. Possible options include "full" corresponding to a model including both a level change and a slope change, "level" corresponding to a model including just a level change, and "slope" corresponding to a model including just a slope change. Default value is "full".
counterfactual	Logical - indicating whether the model-based counterfactual values should also be returned as an additional column in the data. Default value is FALSE, in which case the counterfactual values are not returned.
print_summary	Logical - indicating whethwe the entire model summary should be printed, or just the relevant effect size. Default value is FALSE in which case only the effect size, together with its 95% CI and P-value, are printed.

Value

The function returns a list with three elements: the fitted linear regression model, the summary of the model (including the mean difference and Cohen's d), and the original data together with the model predictions.

Examples

data <- unemployed
form <- as.formula("percent ~ time")
intervention_start_ind <- which(data$year==2020 & data$month>2| data$year==2021)[1]
fit <- its_lm_fourier(data=data,form=form,time_name = "time",intervention_start_ind=intervention_start_ind,freq=12, keep_significant_fourier=TRUE, impact_model = "full",counterfactual = TRUE)

Yael-Travis-Lumer/its2es documentation built on Oct. 31, 2022, 8:05 a.m.