Description Usage Arguments Details Value Author(s) See Also Examples
Subset TAPS data by outcome and covariates, model the attrition rates, impute data for attrited individuals, and find weights for analysis
1 2 3 4 |
interact |
A logical vector indicating if the function is to be run interactively. If TRUE arguments are not needed - default is TRUE |
outcome |
A character vector of the names of outcome variables of interest. It is highly suggested that the outcome variables be entered starting with the earliest wave. |
covars |
A character vector of the names of the covariate variables of interest |
weight |
A logical argument specifying whether to use TAPS base weights or not - default is FALSE |
refusedasNA |
A logical argument specifying whether to conisder the response 'Refused' as a missing value - default and suggested value is TRUE |
method |
A character object indicating type of imputation to be used. hotdeck for hotdeck imputation, multi (default) for mulitple imputation, none for no imputation |
na.delete |
A logical argument specifying whether to eliminate rows with NAs before calculating weights if method chosen is none - default is TRUE. Only set to FALSE if planning to use NA observations. |
m |
A numeric argument specifying number of imputed data sets to produce if using multiple imputation - default is 5 |
pop.base |
A numeric object specifying which CPS data to use as a baseline. 1 is Dec. 2011, 2 is Dec. 2012, 3 is Dec. 2013. Default is 1. |
trunc_at |
A numeric object specifying where to truncate the weights (what should the max weight be?) - default is 5 |
stringsAsFactors |
A logical vector indicating whether non-numeric variables should be factors rather than strings - default is TRUE |
This package is meant to subset The American Panel Survey (TAPS) data by outcome and by covariate variables of interest through the function weightTAPS
.
The subsetting process accounts for respondents attriting from at least one of the waves under analysis, as well as for outcome non-response.
The variables of interest must be entered exactly as named in the TAPS dataframe. See http://taps.wustl.edu/data-archive or use the variablesTAPS
function
to explore the names of the variables by wave. It is important to revise the particular features of each of the variables of interest.
It is strongly suggested that the outcome variables be entered starting with the earliest wave for easier interpretation of the attrition rates.
Other arguments are listed in the help file of the weightTAPS()
function, and must be considered based on the user's needs.
The function can be run in interactive mode by simply running weightTAPS()
. The user must answer the questions based on her needs.
The function weightTAPS()
should be assigned to an object in order to conduct the analysis of TAPS.
weightTAPS()
returns a subset of the complete TAPS dataset that includes only the outcome variable and covariates specified by the user, a set of standard demographics
and a new variable with the corresponding weight for each respondent.
It also retains the respondents that gave an answer to the outcome variable of interest through the waves specified by the user.
Respondents that attritted or did not provide an answer to the outcome variable for any of the waves under analysis are removed from the subset data.
Missing values in sociodemographic variables are imputed for the respondents that remain in the sample, in order to compute their proper weight.
The missing values observed in the covariates of the remaining respondents are imputed through the method selected by the user.
Once the TAPS data is subset, the function calculates weights based on the demographic group membership of the respondents in the final subset.
These weights will be appended to the end of the data frame(s) with the column name new.weights
.
The output (see weightTAPSoutput
) is of class weightTAPSoutput. This class implies the existence of certain slots that save useful information.
These slots are df
, attrit
and stats
.
The slot df
contains the dataframe(s) that represent the final subset data.
The final subset data keeps only the outcome and covariates of interest specified as well as a set of demographics and the new dynamic weights.
It also accounts for non-response in the outcome variable and attrition across waves through the waves specified.
Respondents with missing values in the outcome variable for any of the waves desired are removed from the final dataframe.
The missing covariate data can be left as is, by specifying method="none"
.
If imputation is desired, the argument method
can be set to 'multi' for multiple imputation, or 'hotdeck' for hotdeck imputation.
If multiple imputation is done, the argument m
should be set to the number of imputed dataframes to be created.
Depending on the imputation method selected, df
can be a list of m
elements or a list containing a single element. Each element of df
stores a dataframe.
If method="multi"
is specified, df
contains a list of m
dataframes.
To access the dataframes, use getdf(objectname)
. Objectname corresponds to the object where the value of weightTAPS()
was originally stored.
If hotdeck or no imputation was used, the final dataset is the first element of the df
list, and can be accessed with getdf(objectname)[[1]]
.
The slot attrit
is a list of attrition rates from the first wave specified in the outcome argument.
Each quantity represents the percentage of people (by demographic group) that attritted TAPS through the waves specified.
It compares the initial composition of each demographic group (from the oldest wave specified) to the composition of the same demographic group in the final subset data delivered by weightTAPS()
.
Large values, particularly large values relative to other values in the same sociodemographic category, indicate high rates of attrition.
It is important to highlight that high rates of attrition may cause problems in data analysis.
The slot stats
lists each sociodemographic group's share of the overall population as represented in the final sample for each outcome.
The information contained in both the attrit
and stats
slots can be graphically illustrated using the plot(objectname)
function.
Two different types of plots are displayed after running the plot function: a dot chart and a set of trend plots.
The dot chart shows the differences between the sociodemographic composition of the sample in the first wave specified and the final subset dataframe.
This information is disaggregated by the following sociodemographic groups: Age and Gender, Ethnicity, Education, Income, Region and Metropolitan status, and Internet use.
The trend plots presented illustrate the changing composition of the sample by demographic group across the waves specified.
The lines shown in each plot correspond to the different categories within each of the groups mentioned.
The lines show the percentage of the final subset data belonging to each category by wave.
The plots aim to show the variation in the composition of the sociodemographic groups through the waves specified.
An object of class weightTAPSoutput with the following slots:
df
List of dataframes (or single dataframe) of subsetted TAPS data with imputed data for missing values in covariates, with column for weights
attrit
A list of vectors of attrition rates for populations
stats
A list of population statistics from wave to wave
David Carlson carlson.david@wustl.edu, Michelle Torres smtorres@wustl,and Taeyong Park t.park@wustl.edu
weightTAPSPACK
variablesTAPS
subsetTAPS
weightTAPSoutput
simpleWeight
attritTAPS
multipleImp
hotdeckImp
wavesTAPS
1 2 3 4 5 |
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.