In andrew-leroux/rnhanesdata: NHANES Accelerometry Data Pipeline
knitr::opts_chunk$set(
collapse = TRUE,
comment = "#>"
)
In this vignette we walk through the creation of the 8 primary processed datasets contained in the rnhanesdata package. These
datasets can be organized into four categories:
- Activity count data (accelerometer)
- PAXINTEN_C
- PAXINTEN_D
- Wear/non-wear flags associated (derived from the activity count data)
- Flags_C
- Flags_D
- Mortality data linked to NHANES
- Mortality_2011_C
- Mortality_2011_D
- NHANES demographic, survey design, lifestyle, and comorbiditiy variables
- Covariate_C
- Covariate_D
The _C and _D naming convention indicates which NHANES wave the data is associated with.
More specifically, _C and _D correspond to the 2003-2004 and 2005-2006 waves, respectively.
This document assumes users have some familiarity with the contest of each of these datasets.
See the help files associated with each of these datasets for more information.
Below we show users how to process this data for themselves and then write out the data in a format of their preference.
The first step is to load the rnhanesdata package.
library("rnhanesdata")
Processing the accelerometry data and creating wear/non-wear flags
Using the functions process_accel() and process_flags(), processing the accelerometry data and creating wear/non-wear flags is simple.
In the code below we show how to download the physical activity monitor data from the NHANES website and then calculate wear/non-wear flags.
accel_ls <- process_accel(names_accel_xpt = c("PAXRAW_C","PAXRAW_D"), local=FALSE,
urls=c("https://wwwn.cdc.gov/Nchs/Nhanes/2003-2004/PAXRAW_C.ZIP",
"https://wwwn.cdc.gov/Nchs/Nhanes/2005-2006/PAXRAW_D.ZIP")
)
flags_ls <- process_flags(accel_ls)
This whole process will take some time to finish (approximately 30 minutes on a standard laptop).
Downloading the data directly through R and then unzipping the data in R is, in general, slower
than downloading the data outside of R and then reading in the (un)zipped data locally.
See ?process_accel for details on how to process data that is stored locally.
However, in the interest for full reproducibility we allow for the entire data processing pipeline to be conducted within R.
Once this code has executed, the object accel_ls will be a list where the first element is a data frame corresponding to the 2003-2004 accelerometry data in the 1440+ format. Similarly, the second element of accel_ls will be a data frame containing the 2005-2006 accelerometry data in the 1440+ format. These data frames are
identical to the data objects PAXINTEN_C (2003-2004) and PAXINTEN_D (2005-2006) which are included in the rnhanesdata package.
Type ?PAXINTEN_C and ?PAXINTEN_D in your R console for additional details.
The flags_ls is object will also be a list containing two elements: Flags_C and Flags_D, corresponding the estimated wear/non-wear flag matrices for the 2003-2004 and 2005-2006 accelerometry data, respectively. The format (i.e. column names, dimensions, etc) of Flags_C and Flags_D is identical to that of PAXINTEN_C and PAXINTEN_D, only the activity count data has been replaced by wear/non-wear flags. Type ?Flags_C and ?Flags_D in your R console for additional details.
Saving the data locally is then straightforward. For example, if we wanted to create .rda files which are identical to those included in the package, we would use the following code
PAXINTEN_C <- accel_ls$PAXINTEN_C
PAXINTEN_D <- accel_ls$PAXINTEN_D
Flags_C <- flags_ls$FLags_C
Flags_D <- flags_ls$FLags_D
save(PAXINTEN_C, file="PAXINTEN_C.rda", compress="xz")
save(PAXINTEN_D, file="PAXINTEN_D.rda", compress="xz")
save(Flags_C, file="Flags_C.rda", compress="xz")
save(Flags_D, file="Flags_D.rda", compress="xz")
Changing the argument to file will control where the data are saved.
By default the syntax above will save to wherever the current working directory is, which can be viewed by typing getwd() in the R console.
Note that we use "xz"" compression here. Saving the activity count data using this compression type takes longer to complete than the default compression type, but results in substantially smaller file sizes.
If instead of a .rda file, you'd prefer to save the processed data as .csv, you can use the function write.csv() which uses similar syntax.
write.csv(PAXINTEN_C, file="PAXINTEN_C.csv", row.names=FALSE)
write.csv(PAXINTEN_D, file="PAXINTEN_D.csv", row.names=FALSE)
write.csv(Flags_C, file="Flags_C.csv", row.names=FALSE)
write.csv(Flags_D, file="Flags_D.csv", row.names=FALSE)
Processing the mortality data
Processing the NHANES 2003-2004 and 2005-2006 mortality data is similarly simple using the process_mort() function contained in the rnhanesdata package.
mort_ls <- process_mort()
By default the process_mort() function looks for the raw mortality data within the rnhanesdata package.
This can be changed by supplying a local directory to the
localpath argument. See ?process_mort for more details.
The mortality data can then be saved in the same way that the activity count and wear/non-wear flag data were saved in the previous section
Mortality_2011_C <- mort_ls$Mortality_2011_C
Mortality_2011_D <- mort_ls$Mortality_2011_D
save(Mortality_2011_C, file="Mortality_2011_C.rda", compress="xz")
save(Mortality_2011_D, file="Mortality_2011_D.rda", compress="xz")
Processing the demographic, survey design, lifestyle, and comorbiditiy variables
In this section we will reproduce the Covariate_C and Covariate_D datasets included in the rnhanesdata package.
Processing the demographic, survey design, lifestyle and comorbidity variables included in the Covariate_C and Covariate_D
datasets is a bit more involved than processing the activity count, wear/non-wear flags, and mortality data. First,
the raw data used to create these sets of processed data are spread across multiple SAS XPORT (.xpt) files.
As a result, there's an initial data merging step that needs to take place.
The process_covar() function was created to make this process as straightforward as possible.
It will search a local directory for
all .xpt files associated with a specific NHANES wave. By default process_covar() searches the raw data stored within the rnhanesdata package,
though the directory can be specified by users.
Users can choose to extract all variables across all relevant .xpt files, or request a specific set of variables.
For example, creating the Covariate_C and Covariate_D datasets requires 33 variables.
In addition to the issue of data merging, creating many common variables of interest requires careful consideration of questionnaire skip patterns and
coding for missing data, which can change from wave-to-wave. One of the variables we use to classify individual's alcohol consumption is an example of this
(ALQ130). Accordingly, we process this data separately for each wave.
Below, we show how to extract and merge these 33 variables required to create Covariate_C and Covariate_D .
covar_ls <- process_covar(waves=c("C","D"),
varnames = c("SDDSRVYR","WTMEC2YR", "WTINT2YR",
"SDMVPSU", "SDMVSTRA",
"RIDAGEMN", "RIDAGEEX", "RIDAGEYR", "RIDRETH1", "RIAGENDR",
"BMXWT", "BMXHT", "BMXBMI", "DMDEDUC2",
"ALQ101", "ALQ110", "ALQ120Q","ALQ120U", "ALQ130", "SMQ020", "SMD030", "SMQ040",
"MCQ220","MCQ160F", "MCQ160B", "MCQ160C",
"PFQ049","PFQ054","PFQ057","PFQ059", "PFQ061B", "PFQ061C", "DIQ010"))
Covariate_C <- covar_ls$Covariate_C
Covariate_D <- covar_ls$Covariate_D
The process_covar function will print to the console how many variables were found for each of the NHANES waves considered (see output above).
At this point all variables in each element of covar_ls are numeric, as can be seen by using the str() function
str(Covariate_C)
Many variables of interest are represented by a single NHANES question.
For example, "Gender" can be derived entirely from the NHANES variable "RIAGENDR".
In order to create the "Gender" variable, we need to look at the NHANES demographic data codebook, available at https://wwwn.cdc.gov/Nchs/Nhanes/2003-2004/DEMO_C.htm.
Below, we show how to create the variables included in Covariate_C and Covariate_D which can be derived from a single NHANES question.
Covariate_C$Race <- factor(Covariate_C$RIDRETH1, levels=1:5, labels=c("Mexican American", "Other Hispanic", "White", "Black", "Other"), ordered=FALSE)
Covariate_D$Race <- factor(Covariate_D$RIDRETH1, levels=1:5, labels=c("Mexican American", "Other Hispanic", "White", "Black", "Other"), ordered=FALSE)
Covariate_C$Race <- relevel(Covariate_C$Race, ref="White")
Covariate_D$Race <- relevel(Covariate_D$Race, ref="White")
Covariate_C$Gender <- factor(Covariate_C$RIAGENDR, levels=1:2, labels=c("Male","Female"), ordered=FALSE)
Covariate_D$Gender <- factor(Covariate_D$RIAGENDR, levels=1:2, labels=c("Male","Female"), ordered=FALSE)
Covariate_C$Diabetes <- factor(Covariate_C$DIQ010, levels=c(1,2,3,7,9), labels=c("Yes","No","Borderline","Refused","Don't know"), ordered=FALSE)
Covariate_D$Diabetes <- factor(Covariate_D$DIQ010, levels=c(1,2,3,7,9), labels=c("Yes","No","Borderline","Refused","Don't know"), ordered=FALSE)
Covariate_C$Diabetes <- relevel(Covariate_C$Diabetes, ref="No")
Covariate_D$Diabetes <- relevel(Covariate_D$Diabetes, ref="No")
Covariate_C$CHF <- factor(Covariate_C$MCQ160B, levels=c(1,2,7,9), labels=c("Yes","No","Refused","Don't know"), ordered=FALSE)
Covariate_D$CHF <- factor(Covariate_D$MCQ160B, levels=c(1,2,7,9), labels=c("Yes","No","Refused","Don't know"), ordered=FALSE)
Covariate_C$CHF <- relevel(Covariate_C$CHF, ref="No")
Covariate_D$CHF <- relevel(Covariate_D$CHF, ref="No")
Covariate_C$CHD <- factor(Covariate_C$MCQ160C, levels=c(1,2,7,9), labels=c("Yes","No","Refused","Don't know"), ordered=FALSE)
Covariate_D$CHD <- factor(Covariate_D$MCQ160C, levels=c(1,2,7,9), labels=c("Yes","No","Refused","Don't know"), ordered=FALSE)
Covariate_C$CHD <- relevel(Covariate_C$CHD, ref="No")
Covariate_D$CHD <- relevel(Covariate_D$CHD, ref="No")
Covariate_C$Cancer <- factor(Covariate_C$MCQ220, levels=c(1,2,7,9), labels=c("Yes","No","Refused","Don't know"), ordered=FALSE)
Covariate_D$Cancer <- factor(Covariate_D$MCQ220, levels=c(1,2,7,9), labels=c("Yes","No","Refused","Don't know"), ordered=FALSE)
Covariate_C$Cancer <- relevel(Covariate_C$Cancer, ref="No")
Covariate_D$Cancer <- relevel(Covariate_D$Cancer, ref="No")
Covariate_C$Stroke <- factor(Covariate_C$MCQ160F, levels=c(1,2,7,9), labels=c("Yes","No","Refused","Don't know"), ordered=FALSE)
Covariate_D$Stroke <- factor(Covariate_D$MCQ160F, levels=c(1,2,7,9), labels=c("Yes","No","Refused","Don't know"), ordered=FALSE)
Covariate_C$Cancer <- relevel(Covariate_C$Cancer, ref="No")
Covariate_D$Cancer <- relevel(Covariate_D$Cancer, ref="No")
Covariate_C$EducationAdult <- factor(Covariate_C$DMDEDUC2, levels=c(1,2,3,4,5,7,9),
labels=c("Less than 9th grade","9-11th grade","High school grad/GED or equivalent",
"Some College or AA degree", "College graduate or above","Refused","Don't know"), ordered=FALSE)
Covariate_D$EducationAdult <- factor(Covariate_D$DMDEDUC2, levels=c(1,2,3,4,5,7,9),
labels=c("Less than 9th grade","9-11th grade","High school grad/GED or equivalent",
"Some College or AA degree", "College graduate or above","Refused","Don't know"), ordered=FALSE)
Covariate_C$BMI <- Covariate_C$BMXBMI
Covariate_D$BMI <- Covariate_D$BMXBMI
Covariate_C$BMI_cat <-cut(Covariate_C$BMI, breaks=c(0, 18.5, 25, 30, Inf), labels=c("Underweight","Normal","Overweight","Obese"))
Covariate_D$BMI_cat <-cut(Covariate_D$BMI, breaks=c(0, 18.5, 25, 30, Inf), labels=c("Underweight","Normal","Overweight","Obese"))
Covariate_C$BMI_cat <- relevel(Covariate_C$BMI_cat, ref="Normal")
Covariate_D$BMI_cat <- relevel(Covariate_D$BMI_cat, ref="Normal")
Observe that there are some variables which we applied the relevel() function to. The relevel() function allows us to set the level of the factor variable which is
considered as the "baseline". This is particularly useful when you intend to use factor variables in a regression.
Unlike the variables derived above, some variables (alcohol consumption, cigarette smoking, mobility problem, etc)
require integrating the information from multiple variables (multiple questionnaire items).
Often, creating variables derived from more than one questionnaire item require choices involving how the definition of said variables.
Moreover, creating these variables requires careful attention to "skip patterns" in the relevant questionnaires in order to not accidentally mark data as "missing" when
there is really an implied value based on their responses to other questions.
Creating a variable for cigarette smoking
The NHANES 2003-2004 and 2005-2006 waves ask a number of questions about smoking.
A very simple way to classify individuals based on their smoking history is:
- Never smoker
- Former smoker
- Current smoker
However, NHANES does not ask a question that corresponds exactly to this classification.
Instead, the first question on the smoking questionnaire for these two waves is: "Have you smoked at least 100 cigarettes in your entire life"?
If the answer to this question is "no", "refused", or "don't know", then the participant is asked no more questions about their cigarette smoking history.
If the answer to this question is "yes" then they're asked at least 2 more questions on smoking history.
There are many possible ways to categorize individuals into never, former, or current smokers using the smoking questionnaire from NHANES 2003-2004/2005-2006.
In the Covariate_C and Covariate_D data sets, we only consider individuals' responses to the questions:
- SMQ020: "Have you smoked at least 100 cigarettes in your entire life"
- SMQ040: "Do you now smoke cigarettes"
If an individual responded "No" to question 1 above, we classify them as Never smokers.
If an individual responds "Yes" to question 1 and "Not at all" to question 2, we classify them as former smokers.
If an individual response "Yes" to question 1 and either "Every day" or "Some days". Any other combination of responses is considered missing data.
The code to create our cigarette smoking variable is then
## temporarily assign the SmokeCigs variable as individuals'
## response to SMQ040
Covariate_C$SmokeCigs <-Covariate_C$SMQ040
Covariate_D$SmokeCigs <-Covariate_D$SMQ040
## re-codes individuals who responded "No" to "Ever smoked 100 cigarettes in your life" (SMQ020)
## as -999 instead of missing (since these individuals were never asked question SMQ040)
Covariate_C$SmokeCigs[Covariate_C$SMQ020 == 2] <- -999
Covariate_D$SmokeCigs[Covariate_D$SMQ020 == 2] <- -999
## re-code individuals who answered "some days" to SMQ040 ("do you now smoke cigarettes")
## as "1". These individuals are considered current smokers by our definition
Covariate_C$SmokeCigs[Covariate_C$SmokeCigs == 2] <- 1
Covariate_D$SmokeCigs[Covariate_D$SmokeCigs == 2] <- 1
## finally, create the factor variable based on our re-coding
Covariate_C$SmokeCigs <- factor(Covariate_C$SmokeCigs, levels=c(-999, 3, 1), labels=c("Never", "Former", "Current"), ordered=FALSE)
Covariate_D$SmokeCigs <- factor(Covariate_D$SmokeCigs, levels=c(-999, 3, 1), labels=c("Never", "Former", "Current"), ordered=FALSE)
Creating a variable for alcohol consumption
The NHANES 2003-2004 and 2005-2006 waves ask a number of questions about alcohol consumption.
Similar to how we classified individuals into smoking categories, it's common to classify individuals as being one of
- Non drinker
- Moderate drinker
- Heavy drinker
The CDC provides a (gender specific) definition for moderate drinker versus heavy drinker based on the amount of alcohol consumed, on average, per day AND whether
or not an individual engages in binge drinking. However, as of writing, the NHANES 2003-2006 data do not contain data on binge drinking as used in
the CDC's definition of heavy drinker. As such, we use individuals' self-reported number of drinks per week to classify individuals.
We define non drinkers as anyone who reported never drinking 12 or more drinks over the course of their entire life (ALQ101), or over the course of any year (ALQ110).
We defined former drinkers as anyone who indicated that they had at least 12 drinks over the course of any one year, but report drinking no alcohol in the past 12 months.
Moderate drinkers are women and men who report drinking, on average, 7 and 14 fewer drinks per week, respectively. Anyone who does not have sufficient
data to identify them as belonging to one of these groups was considered to have missing data.
There is a slight complication in that the questionnaire is structured such that individuals report how often they drank alcohol over the past year (ALQ120Q)
among individuals who reported drinking alcohol, the number of days they drank per unit time (ALQ120U: week, month, year), and then how many drinks
they did drink on those days where they consumed alcohol (ALQ130). The answers to these three questions need to be considered jointly in order to calculate a variable
corresponding to the "average number of alcoholic drinks drank per week over the last 12 months" variable. An additional complication is that
NHANES changed how refused/missing data were coded for the variable ALQ130 (# of drinks on those days that alcohol was drunk) this is addressed in the code below.
## classifies don't know/refused as missing
Covariate_C$ALQ101[Covariate_C$ALQ101 %in% c(7,9)] <- NA
Covariate_D$ALQ101[Covariate_D$ALQ101 %in% c(7,9)] <- NA
Covariate_C$ALQ110[Covariate_C$ALQ110 %in% c(7,9)] <- NA
Covariate_D$ALQ110[Covariate_D$ALQ110 %in% c(7,9)] <- NA
## get a factor variable which corresponds to "have you ever in your life had 12 drinks total over the course of a year?"
Covariate_C$Alcohol_Ever <- factor(as.numeric(Covariate_C$ALQ101 == 1 | Covariate_C$ALQ110 == 1), levels=c(1,0), labels=c("Yes","No"), ordered=FALSE)
Covariate_D$Alcohol_Ever <- factor(as.numeric(Covariate_D$ALQ101 == 1 | Covariate_D$ALQ110 == 1), levels=c(1,0), labels=c("Yes","No"), ordered=FALSE)
## re-code "how often drink alcohol over past 12 mos" = refused (777) and don't know (999) as missing
Covariate_C$ALQ120Q[Covariate_C$ALQ120Q %in% c(777,999)] <- NA
Covariate_D$ALQ120Q[Covariate_D$ALQ120Q %in% c(777,999)] <- NA
## re code # days drank alcohol units of refused/don't know as missing ()
## note: there are no observed values of 7/9 in these variables, but they are options in the survey
Covariate_C$ALQ120U[Covariate_C$ALQ120U %in% c(7,9)] <- NA
Covariate_D$ALQ120U[Covariate_D$ALQ120U %in% c(7,9)] <- NA
## re code # of drinks on those days that alcohol was drunk
## to be NA where the answer was "refused" or "dont know"
## note they changed the coding between 2003-2004 and 2005 and 2006 waves from 77/99 to 777/999
Covariate_C$ALQ130[Covariate_C$ALQ130 %in% c(77,99)] <- NA
Covariate_D$ALQ130[Covariate_D$ALQ130 %in% c(777,999)] <- NA
## get number of drinks per week for all individuals
multiplier <- 7*c(1/7, 1/30, 1/365)
## (#days drank alcohol / unit time) * (unit time / week) * (drinks / day) = drinks/week
Covariate_C$DrinksPerWeek <- Covariate_C$ALQ120Q * (multiplier[Covariate_C$ALQ120U]) * Covariate_C$ALQ130
Covariate_D$DrinksPerWeek <- Covariate_D$ALQ120Q * (multiplier[Covariate_D$ALQ120U]) * Covariate_D$ALQ130
## recode individuals who are never drinkers OR non-drinkers as drinking 0 drinks per week
Covariate_C$DrinksPerWeek[Covariate_C$Alcohol_Ever == "No"] <- 0
Covariate_D$DrinksPerWeek[Covariate_D$Alcohol_Ever == "No"] <- 0
Covariate_C$DrinksPerWeek[Covariate_C$ALQ120Q == 0] <- 0
Covariate_D$DrinksPerWeek[Covariate_D$ALQ120Q == 0] <- 0
## classify individuals as "non-drinker","moderate","heavy" using gender specific CDC thresholds of
## no more than 7 drinks/week for women, and no more than 14 drinks per week for men.
## note we do not have
cutoff <- c(14,7)
Covariate_C$DrinkStatus <- cut(Covariate_C$DrinksPerWeek / cutoff[Covariate_C$RIAGENDR], breaks=c(-1, 0, 1, Inf),
labels=c("Non-Drinker","Moderate Drinker", "Heavy Drinker"))
Covariate_D$DrinkStatus <- cut(Covariate_D$DrinksPerWeek / cutoff[Covariate_D$RIAGENDR], breaks=c(-1, 0, 1, Inf),
labels=c("Non-Drinker","Moderate Drinker", "Heavy Drinker"))
## re-level the factor variable to have the baseline be moderate drinkers
Covariate_C$DrinkStatus <- relevel(Covariate_C$DrinkStatus, ref="Moderate Drinker")
Covariate_D$DrinkStatus <- relevel(Covariate_D$DrinkStatus, ref="Moderate Drinker")
Creating a variable for mobility problem
We define participants as reporting a mobility problem if they report any of:
- Difficulty walking a quarter mile (PFQ061B)
- Difficulty climbing 10 stairs (PFQ061C)
- Use of any special equipment to walk (PFQ054)
The skip pattern of the physical function questionnaire is crucial to correctly creating this variable.
That is, if individuals report needing special equipment to walk, they are not asked about their difficulty climbing stairs or walking a quarter mile.
In addition, individuals are not asked about difficulty walking or climbing stairs if they
are 59 or younger according the variable RIDAGEYR (not RIDAGEMN which measures age at interview) and indicate they have no limitations that keep them
from working (PFQ049), no memory problems (PFQ057), and no limitations arising from a mental, emotional, or physical problem (PFQ059).
We consider these individuals to have no mobility problem.
Covariate_C$Difficulty_Walking <- factor(as.numeric(Covariate_C$PFQ061B==1), levels=c(1,0), labels=c("No Difficulty","Any Difficulty"), ordered=FALSE)
Covariate_D$Difficulty_Walking <- factor(as.numeric(Covariate_D$PFQ061B==1), levels=c(1,0), labels=c("No Difficulty","Any Difficulty"), ordered=FALSE)
Covariate_C$Difficulty_Stairs <- factor(as.numeric(Covariate_C$PFQ061C==1), levels=c(1,0), labels=c("No Difficulty","Any Difficulty"), ordered=FALSE)
Covariate_D$Difficulty_Stairs <- factor(as.numeric(Covariate_D$PFQ061C==1), levels=c(1,0), labels=c("No Difficulty","Any Difficulty"), ordered=FALSE)
## label anyone who requires special equipment to walk as "any difficulty"
inx_sp_equip_C <- which(Covariate_C$PFQ054 == 1)
inx_sp_equip_D <- which(Covariate_D$PFQ054 == 1)
Covariate_C$Difficulty_Walking[inx_sp_equip_C] <- "Any Difficulty"
Covariate_D$Difficulty_Walking[inx_sp_equip_D] <- "Any Difficulty"
Covariate_C$Difficulty_Stairs[inx_sp_equip_C] <- "Any Difficulty"
Covariate_D$Difficulty_Stairs[inx_sp_equip_D] <- "Any Difficulty"
rm(list=c("inx_sp_equip_C","inx_sp_equip_D"))
# label anyone 59 and younger at interview who responds no to PFQ049, PFQ057, PFQ059 as "No difficulty"
inx_good_fn_C <- which(Covariate_C$PFQ049 == 2 & Covariate_C$PFQ057 == 2 & Covariate_C$PFQ059 == 2 & Covariate_C$RIDAGEYR <= 59)
inx_good_fn_D <- which(Covariate_D$PFQ049 == 2 & Covariate_D$PFQ057 == 2 & Covariate_D$PFQ059 == 2 & Covariate_D$RIDAGEYR <= 59)
Covariate_C$Difficulty_Walking[inx_good_fn_C] <- "No Difficulty"
Covariate_D$Difficulty_Walking[inx_good_fn_D] <- "No Difficulty"
Covariate_C$Difficulty_Stairs[inx_good_fn_C] <- "No Difficulty"
Covariate_D$Difficulty_Stairs[inx_good_fn_D] <- "No Difficulty"
Covariate_C$MobilityProblem <- factor(as.numeric(Covariate_C$Difficulty_Stairs == "Any Difficulty" | Covariate_C$Difficulty_Walking == "Any Difficulty"),
levels=c(0,1), labels=c("No Difficulty","Any Difficulty"),ordered=FALSE)
Covariate_D$MobilityProblem <- factor(as.numeric(Covariate_D$Difficulty_Stairs == "Any Difficulty" | Covariate_D$Difficulty_Walking == "Any Difficulty"),
levels=c(0,1), labels=c("No Difficulty","Any Difficulty"),ordered=FALSE)
Saving the demographic, lifestyle, and comorbidity data
Finally, we save the columns of interest from these data files
vars_inc <- c("SEQN","SDDSRVYR",
"SDMVPSU","SDMVSTRA","WTINT2YR","WTMEC2YR",
"RIDAGEMN","RIDAGEEX","RIDAGEYR","BMI","BMI_cat","Race","Gender",
"Diabetes","CHF","CHD","Cancer","Stroke",
"EducationAdult","MobilityProblem",
"DrinkStatus","DrinksPerWeek","SmokeCigs")
Covariate_C <- Covariate_C[,vars_inc]
Covariate_D <- Covariate_D[,vars_inc]
save(Covariate_C, file="Covariate_C.rda", compress="xz")
save(Covariate_D, file="Covariate_D.rda", compress="xz")
andrew-leroux/rnhanesdata documentation built on March 6, 2020, 11:35 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
knitr::opts_chunk$set( collapse = TRUE, comment = "#>" )
In this vignette we walk through the creation of the 8 primary processed datasets contained in the rnhanesdata package. These datasets can be organized into four categories:
- Activity count data (accelerometer)
- PAXINTEN_C
- PAXINTEN_D
- Wear/non-wear flags associated (derived from the activity count data)
- Flags_C
- Flags_D
- Mortality data linked to NHANES
- Mortality_2011_C
- Mortality_2011_D
- NHANES demographic, survey design, lifestyle, and comorbiditiy variables
- Covariate_C
- Covariate_D
The _C and _D naming convention indicates which NHANES wave the data is associated with. More specifically, _C and _D correspond to the 2003-2004 and 2005-2006 waves, respectively. This document assumes users have some familiarity with the contest of each of these datasets. See the help files associated with each of these datasets for more information.
Below we show users how to process this data for themselves and then write out the data in a format of their preference. The first step is to load the rnhanesdata package.
library("rnhanesdata")
Processing the accelerometry data and creating wear/non-wear flags
Using the functions process_accel() and process_flags(), processing the accelerometry data and creating wear/non-wear flags is simple. In the code below we show how to download the physical activity monitor data from the NHANES website and then calculate wear/non-wear flags.
accel_ls <- process_accel(names_accel_xpt = c("PAXRAW_C","PAXRAW_D"), local=FALSE, urls=c("https://wwwn.cdc.gov/Nchs/Nhanes/2003-2004/PAXRAW_C.ZIP", "https://wwwn.cdc.gov/Nchs/Nhanes/2005-2006/PAXRAW_D.ZIP") ) flags_ls <- process_flags(accel_ls)
This whole process will take some time to finish (approximately 30 minutes on a standard laptop). Downloading the data directly through R and then unzipping the data in R is, in general, slower than downloading the data outside of R and then reading in the (un)zipped data locally. See ?process_accel for details on how to process data that is stored locally. However, in the interest for full reproducibility we allow for the entire data processing pipeline to be conducted within R.
Once this code has executed, the object accel_ls will be a list where the first element is a data frame corresponding to the 2003-2004 accelerometry data in the 1440+ format. Similarly, the second element of accel_ls will be a data frame containing the 2005-2006 accelerometry data in the 1440+ format. These data frames are identical to the data objects PAXINTEN_C (2003-2004) and PAXINTEN_D (2005-2006) which are included in the rnhanesdata package. Type ?PAXINTEN_C and ?PAXINTEN_D in your R console for additional details. The flags_ls is object will also be a list containing two elements: Flags_C and Flags_D, corresponding the estimated wear/non-wear flag matrices for the 2003-2004 and 2005-2006 accelerometry data, respectively. The format (i.e. column names, dimensions, etc) of Flags_C and Flags_D is identical to that of PAXINTEN_C and PAXINTEN_D, only the activity count data has been replaced by wear/non-wear flags. Type ?Flags_C and ?Flags_D in your R console for additional details.
Saving the data locally is then straightforward. For example, if we wanted to create .rda files which are identical to those included in the package, we would use the following code
PAXINTEN_C <- accel_ls$PAXINTEN_C PAXINTEN_D <- accel_ls$PAXINTEN_D Flags_C <- flags_ls$FLags_C Flags_D <- flags_ls$FLags_D save(PAXINTEN_C, file="PAXINTEN_C.rda", compress="xz") save(PAXINTEN_D, file="PAXINTEN_D.rda", compress="xz") save(Flags_C, file="Flags_C.rda", compress="xz") save(Flags_D, file="Flags_D.rda", compress="xz")
Changing the argument to file will control where the data are saved. By default the syntax above will save to wherever the current working directory is, which can be viewed by typing getwd() in the R console. Note that we use "xz"" compression here. Saving the activity count data using this compression type takes longer to complete than the default compression type, but results in substantially smaller file sizes.
If instead of a .rda file, you'd prefer to save the processed data as .csv, you can use the function write.csv() which uses similar syntax.
write.csv(PAXINTEN_C, file="PAXINTEN_C.csv", row.names=FALSE) write.csv(PAXINTEN_D, file="PAXINTEN_D.csv", row.names=FALSE) write.csv(Flags_C, file="Flags_C.csv", row.names=FALSE) write.csv(Flags_D, file="Flags_D.csv", row.names=FALSE)
Processing the mortality data
Processing the NHANES 2003-2004 and 2005-2006 mortality data is similarly simple using the process_mort() function contained in the rnhanesdata package.
mort_ls <- process_mort()
By default the process_mort() function looks for the raw mortality data within the rnhanesdata package. This can be changed by supplying a local directory to the localpath argument. See ?process_mort for more details.
The mortality data can then be saved in the same way that the activity count and wear/non-wear flag data were saved in the previous section
Mortality_2011_C <- mort_ls$Mortality_2011_C Mortality_2011_D <- mort_ls$Mortality_2011_D save(Mortality_2011_C, file="Mortality_2011_C.rda", compress="xz") save(Mortality_2011_D, file="Mortality_2011_D.rda", compress="xz")
Processing the demographic, survey design, lifestyle, and comorbiditiy variables
In this section we will reproduce the Covariate_C and Covariate_D datasets included in the rnhanesdata package. Processing the demographic, survey design, lifestyle and comorbidity variables included in the Covariate_C and Covariate_D datasets is a bit more involved than processing the activity count, wear/non-wear flags, and mortality data. First, the raw data used to create these sets of processed data are spread across multiple SAS XPORT (.xpt) files. As a result, there's an initial data merging step that needs to take place. The process_covar() function was created to make this process as straightforward as possible. It will search a local directory for all .xpt files associated with a specific NHANES wave. By default process_covar() searches the raw data stored within the rnhanesdata package, though the directory can be specified by users. Users can choose to extract all variables across all relevant .xpt files, or request a specific set of variables. For example, creating the Covariate_C and Covariate_D datasets requires 33 variables.
In addition to the issue of data merging, creating many common variables of interest requires careful consideration of questionnaire skip patterns and coding for missing data, which can change from wave-to-wave. One of the variables we use to classify individual's alcohol consumption is an example of this (ALQ130). Accordingly, we process this data separately for each wave.
Below, we show how to extract and merge these 33 variables required to create Covariate_C and Covariate_D .
covar_ls <- process_covar(waves=c("C","D"), varnames = c("SDDSRVYR","WTMEC2YR", "WTINT2YR", "SDMVPSU", "SDMVSTRA", "RIDAGEMN", "RIDAGEEX", "RIDAGEYR", "RIDRETH1", "RIAGENDR", "BMXWT", "BMXHT", "BMXBMI", "DMDEDUC2", "ALQ101", "ALQ110", "ALQ120Q","ALQ120U", "ALQ130", "SMQ020", "SMD030", "SMQ040", "MCQ220","MCQ160F", "MCQ160B", "MCQ160C", "PFQ049","PFQ054","PFQ057","PFQ059", "PFQ061B", "PFQ061C", "DIQ010")) Covariate_C <- covar_ls$Covariate_C Covariate_D <- covar_ls$Covariate_D
The process_covar function will print to the console how many variables were found for each of the NHANES waves considered (see output above). At this point all variables in each element of covar_ls are numeric, as can be seen by using the str() function
str(Covariate_C)
Many variables of interest are represented by a single NHANES question. For example, "Gender" can be derived entirely from the NHANES variable "RIAGENDR". In order to create the "Gender" variable, we need to look at the NHANES demographic data codebook, available at https://wwwn.cdc.gov/Nchs/Nhanes/2003-2004/DEMO_C.htm. Below, we show how to create the variables included in Covariate_C and Covariate_D which can be derived from a single NHANES question.
Covariate_C$Race <- factor(Covariate_C$RIDRETH1, levels=1:5, labels=c("Mexican American", "Other Hispanic", "White", "Black", "Other"), ordered=FALSE) Covariate_D$Race <- factor(Covariate_D$RIDRETH1, levels=1:5, labels=c("Mexican American", "Other Hispanic", "White", "Black", "Other"), ordered=FALSE) Covariate_C$Race <- relevel(Covariate_C$Race, ref="White") Covariate_D$Race <- relevel(Covariate_D$Race, ref="White") Covariate_C$Gender <- factor(Covariate_C$RIAGENDR, levels=1:2, labels=c("Male","Female"), ordered=FALSE) Covariate_D$Gender <- factor(Covariate_D$RIAGENDR, levels=1:2, labels=c("Male","Female"), ordered=FALSE) Covariate_C$Diabetes <- factor(Covariate_C$DIQ010, levels=c(1,2,3,7,9), labels=c("Yes","No","Borderline","Refused","Don't know"), ordered=FALSE) Covariate_D$Diabetes <- factor(Covariate_D$DIQ010, levels=c(1,2,3,7,9), labels=c("Yes","No","Borderline","Refused","Don't know"), ordered=FALSE) Covariate_C$Diabetes <- relevel(Covariate_C$Diabetes, ref="No") Covariate_D$Diabetes <- relevel(Covariate_D$Diabetes, ref="No") Covariate_C$CHF <- factor(Covariate_C$MCQ160B, levels=c(1,2,7,9), labels=c("Yes","No","Refused","Don't know"), ordered=FALSE) Covariate_D$CHF <- factor(Covariate_D$MCQ160B, levels=c(1,2,7,9), labels=c("Yes","No","Refused","Don't know"), ordered=FALSE) Covariate_C$CHF <- relevel(Covariate_C$CHF, ref="No") Covariate_D$CHF <- relevel(Covariate_D$CHF, ref="No") Covariate_C$CHD <- factor(Covariate_C$MCQ160C, levels=c(1,2,7,9), labels=c("Yes","No","Refused","Don't know"), ordered=FALSE) Covariate_D$CHD <- factor(Covariate_D$MCQ160C, levels=c(1,2,7,9), labels=c("Yes","No","Refused","Don't know"), ordered=FALSE) Covariate_C$CHD <- relevel(Covariate_C$CHD, ref="No") Covariate_D$CHD <- relevel(Covariate_D$CHD, ref="No") Covariate_C$Cancer <- factor(Covariate_C$MCQ220, levels=c(1,2,7,9), labels=c("Yes","No","Refused","Don't know"), ordered=FALSE) Covariate_D$Cancer <- factor(Covariate_D$MCQ220, levels=c(1,2,7,9), labels=c("Yes","No","Refused","Don't know"), ordered=FALSE) Covariate_C$Cancer <- relevel(Covariate_C$Cancer, ref="No") Covariate_D$Cancer <- relevel(Covariate_D$Cancer, ref="No") Covariate_C$Stroke <- factor(Covariate_C$MCQ160F, levels=c(1,2,7,9), labels=c("Yes","No","Refused","Don't know"), ordered=FALSE) Covariate_D$Stroke <- factor(Covariate_D$MCQ160F, levels=c(1,2,7,9), labels=c("Yes","No","Refused","Don't know"), ordered=FALSE) Covariate_C$Cancer <- relevel(Covariate_C$Cancer, ref="No") Covariate_D$Cancer <- relevel(Covariate_D$Cancer, ref="No") Covariate_C$EducationAdult <- factor(Covariate_C$DMDEDUC2, levels=c(1,2,3,4,5,7,9), labels=c("Less than 9th grade","9-11th grade","High school grad/GED or equivalent", "Some College or AA degree", "College graduate or above","Refused","Don't know"), ordered=FALSE) Covariate_D$EducationAdult <- factor(Covariate_D$DMDEDUC2, levels=c(1,2,3,4,5,7,9), labels=c("Less than 9th grade","9-11th grade","High school grad/GED or equivalent", "Some College or AA degree", "College graduate or above","Refused","Don't know"), ordered=FALSE) Covariate_C$BMI <- Covariate_C$BMXBMI Covariate_D$BMI <- Covariate_D$BMXBMI Covariate_C$BMI_cat <-cut(Covariate_C$BMI, breaks=c(0, 18.5, 25, 30, Inf), labels=c("Underweight","Normal","Overweight","Obese")) Covariate_D$BMI_cat <-cut(Covariate_D$BMI, breaks=c(0, 18.5, 25, 30, Inf), labels=c("Underweight","Normal","Overweight","Obese")) Covariate_C$BMI_cat <- relevel(Covariate_C$BMI_cat, ref="Normal") Covariate_D$BMI_cat <- relevel(Covariate_D$BMI_cat, ref="Normal")
Observe that there are some variables which we applied the relevel() function to. The relevel() function allows us to set the level of the factor variable which is considered as the "baseline". This is particularly useful when you intend to use factor variables in a regression.
Unlike the variables derived above, some variables (alcohol consumption, cigarette smoking, mobility problem, etc) require integrating the information from multiple variables (multiple questionnaire items). Often, creating variables derived from more than one questionnaire item require choices involving how the definition of said variables. Moreover, creating these variables requires careful attention to "skip patterns" in the relevant questionnaires in order to not accidentally mark data as "missing" when there is really an implied value based on their responses to other questions.
Creating a variable for cigarette smoking
The NHANES 2003-2004 and 2005-2006 waves ask a number of questions about smoking. A very simple way to classify individuals based on their smoking history is:
- Never smoker
- Former smoker
- Current smoker
However, NHANES does not ask a question that corresponds exactly to this classification. Instead, the first question on the smoking questionnaire for these two waves is: "Have you smoked at least 100 cigarettes in your entire life"? If the answer to this question is "no", "refused", or "don't know", then the participant is asked no more questions about their cigarette smoking history. If the answer to this question is "yes" then they're asked at least 2 more questions on smoking history.
There are many possible ways to categorize individuals into never, former, or current smokers using the smoking questionnaire from NHANES 2003-2004/2005-2006. In the Covariate_C and Covariate_D data sets, we only consider individuals' responses to the questions:
- SMQ020: "Have you smoked at least 100 cigarettes in your entire life"
- SMQ040: "Do you now smoke cigarettes"
If an individual responded "No" to question 1 above, we classify them as Never smokers. If an individual responds "Yes" to question 1 and "Not at all" to question 2, we classify them as former smokers. If an individual response "Yes" to question 1 and either "Every day" or "Some days". Any other combination of responses is considered missing data.
The code to create our cigarette smoking variable is then
## temporarily assign the SmokeCigs variable as individuals' ## response to SMQ040 Covariate_C$SmokeCigs <-Covariate_C$SMQ040 Covariate_D$SmokeCigs <-Covariate_D$SMQ040 ## re-codes individuals who responded "No" to "Ever smoked 100 cigarettes in your life" (SMQ020) ## as -999 instead of missing (since these individuals were never asked question SMQ040) Covariate_C$SmokeCigs[Covariate_C$SMQ020 == 2] <- -999 Covariate_D$SmokeCigs[Covariate_D$SMQ020 == 2] <- -999 ## re-code individuals who answered "some days" to SMQ040 ("do you now smoke cigarettes") ## as "1". These individuals are considered current smokers by our definition Covariate_C$SmokeCigs[Covariate_C$SmokeCigs == 2] <- 1 Covariate_D$SmokeCigs[Covariate_D$SmokeCigs == 2] <- 1 ## finally, create the factor variable based on our re-coding Covariate_C$SmokeCigs <- factor(Covariate_C$SmokeCigs, levels=c(-999, 3, 1), labels=c("Never", "Former", "Current"), ordered=FALSE) Covariate_D$SmokeCigs <- factor(Covariate_D$SmokeCigs, levels=c(-999, 3, 1), labels=c("Never", "Former", "Current"), ordered=FALSE)
Creating a variable for alcohol consumption
The NHANES 2003-2004 and 2005-2006 waves ask a number of questions about alcohol consumption. Similar to how we classified individuals into smoking categories, it's common to classify individuals as being one of
- Non drinker
- Moderate drinker
- Heavy drinker
The CDC provides a (gender specific) definition for moderate drinker versus heavy drinker based on the amount of alcohol consumed, on average, per day AND whether or not an individual engages in binge drinking. However, as of writing, the NHANES 2003-2006 data do not contain data on binge drinking as used in the CDC's definition of heavy drinker. As such, we use individuals' self-reported number of drinks per week to classify individuals.
We define non drinkers as anyone who reported never drinking 12 or more drinks over the course of their entire life (ALQ101), or over the course of any year (ALQ110). We defined former drinkers as anyone who indicated that they had at least 12 drinks over the course of any one year, but report drinking no alcohol in the past 12 months. Moderate drinkers are women and men who report drinking, on average, 7 and 14 fewer drinks per week, respectively. Anyone who does not have sufficient data to identify them as belonging to one of these groups was considered to have missing data.
There is a slight complication in that the questionnaire is structured such that individuals report how often they drank alcohol over the past year (ALQ120Q) among individuals who reported drinking alcohol, the number of days they drank per unit time (ALQ120U: week, month, year), and then how many drinks they did drink on those days where they consumed alcohol (ALQ130). The answers to these three questions need to be considered jointly in order to calculate a variable corresponding to the "average number of alcoholic drinks drank per week over the last 12 months" variable. An additional complication is that NHANES changed how refused/missing data were coded for the variable ALQ130 (# of drinks on those days that alcohol was drunk) this is addressed in the code below.
## classifies don't know/refused as missing Covariate_C$ALQ101[Covariate_C$ALQ101 %in% c(7,9)] <- NA Covariate_D$ALQ101[Covariate_D$ALQ101 %in% c(7,9)] <- NA Covariate_C$ALQ110[Covariate_C$ALQ110 %in% c(7,9)] <- NA Covariate_D$ALQ110[Covariate_D$ALQ110 %in% c(7,9)] <- NA ## get a factor variable which corresponds to "have you ever in your life had 12 drinks total over the course of a year?" Covariate_C$Alcohol_Ever <- factor(as.numeric(Covariate_C$ALQ101 == 1 | Covariate_C$ALQ110 == 1), levels=c(1,0), labels=c("Yes","No"), ordered=FALSE) Covariate_D$Alcohol_Ever <- factor(as.numeric(Covariate_D$ALQ101 == 1 | Covariate_D$ALQ110 == 1), levels=c(1,0), labels=c("Yes","No"), ordered=FALSE) ## re-code "how often drink alcohol over past 12 mos" = refused (777) and don't know (999) as missing Covariate_C$ALQ120Q[Covariate_C$ALQ120Q %in% c(777,999)] <- NA Covariate_D$ALQ120Q[Covariate_D$ALQ120Q %in% c(777,999)] <- NA ## re code # days drank alcohol units of refused/don't know as missing () ## note: there are no observed values of 7/9 in these variables, but they are options in the survey Covariate_C$ALQ120U[Covariate_C$ALQ120U %in% c(7,9)] <- NA Covariate_D$ALQ120U[Covariate_D$ALQ120U %in% c(7,9)] <- NA ## re code # of drinks on those days that alcohol was drunk ## to be NA where the answer was "refused" or "dont know" ## note they changed the coding between 2003-2004 and 2005 and 2006 waves from 77/99 to 777/999 Covariate_C$ALQ130[Covariate_C$ALQ130 %in% c(77,99)] <- NA Covariate_D$ALQ130[Covariate_D$ALQ130 %in% c(777,999)] <- NA ## get number of drinks per week for all individuals multiplier <- 7*c(1/7, 1/30, 1/365) ## (#days drank alcohol / unit time) * (unit time / week) * (drinks / day) = drinks/week Covariate_C$DrinksPerWeek <- Covariate_C$ALQ120Q * (multiplier[Covariate_C$ALQ120U]) * Covariate_C$ALQ130 Covariate_D$DrinksPerWeek <- Covariate_D$ALQ120Q * (multiplier[Covariate_D$ALQ120U]) * Covariate_D$ALQ130 ## recode individuals who are never drinkers OR non-drinkers as drinking 0 drinks per week Covariate_C$DrinksPerWeek[Covariate_C$Alcohol_Ever == "No"] <- 0 Covariate_D$DrinksPerWeek[Covariate_D$Alcohol_Ever == "No"] <- 0 Covariate_C$DrinksPerWeek[Covariate_C$ALQ120Q == 0] <- 0 Covariate_D$DrinksPerWeek[Covariate_D$ALQ120Q == 0] <- 0 ## classify individuals as "non-drinker","moderate","heavy" using gender specific CDC thresholds of ## no more than 7 drinks/week for women, and no more than 14 drinks per week for men. ## note we do not have cutoff <- c(14,7) Covariate_C$DrinkStatus <- cut(Covariate_C$DrinksPerWeek / cutoff[Covariate_C$RIAGENDR], breaks=c(-1, 0, 1, Inf), labels=c("Non-Drinker","Moderate Drinker", "Heavy Drinker")) Covariate_D$DrinkStatus <- cut(Covariate_D$DrinksPerWeek / cutoff[Covariate_D$RIAGENDR], breaks=c(-1, 0, 1, Inf), labels=c("Non-Drinker","Moderate Drinker", "Heavy Drinker")) ## re-level the factor variable to have the baseline be moderate drinkers Covariate_C$DrinkStatus <- relevel(Covariate_C$DrinkStatus, ref="Moderate Drinker") Covariate_D$DrinkStatus <- relevel(Covariate_D$DrinkStatus, ref="Moderate Drinker")
Creating a variable for mobility problem
We define participants as reporting a mobility problem if they report any of:
- Difficulty walking a quarter mile (PFQ061B)
- Difficulty climbing 10 stairs (PFQ061C)
- Use of any special equipment to walk (PFQ054)
The skip pattern of the physical function questionnaire is crucial to correctly creating this variable. That is, if individuals report needing special equipment to walk, they are not asked about their difficulty climbing stairs or walking a quarter mile. In addition, individuals are not asked about difficulty walking or climbing stairs if they are 59 or younger according the variable RIDAGEYR (not RIDAGEMN which measures age at interview) and indicate they have no limitations that keep them from working (PFQ049), no memory problems (PFQ057), and no limitations arising from a mental, emotional, or physical problem (PFQ059). We consider these individuals to have no mobility problem.
Covariate_C$Difficulty_Walking <- factor(as.numeric(Covariate_C$PFQ061B==1), levels=c(1,0), labels=c("No Difficulty","Any Difficulty"), ordered=FALSE) Covariate_D$Difficulty_Walking <- factor(as.numeric(Covariate_D$PFQ061B==1), levels=c(1,0), labels=c("No Difficulty","Any Difficulty"), ordered=FALSE) Covariate_C$Difficulty_Stairs <- factor(as.numeric(Covariate_C$PFQ061C==1), levels=c(1,0), labels=c("No Difficulty","Any Difficulty"), ordered=FALSE) Covariate_D$Difficulty_Stairs <- factor(as.numeric(Covariate_D$PFQ061C==1), levels=c(1,0), labels=c("No Difficulty","Any Difficulty"), ordered=FALSE) ## label anyone who requires special equipment to walk as "any difficulty" inx_sp_equip_C <- which(Covariate_C$PFQ054 == 1) inx_sp_equip_D <- which(Covariate_D$PFQ054 == 1) Covariate_C$Difficulty_Walking[inx_sp_equip_C] <- "Any Difficulty" Covariate_D$Difficulty_Walking[inx_sp_equip_D] <- "Any Difficulty" Covariate_C$Difficulty_Stairs[inx_sp_equip_C] <- "Any Difficulty" Covariate_D$Difficulty_Stairs[inx_sp_equip_D] <- "Any Difficulty" rm(list=c("inx_sp_equip_C","inx_sp_equip_D")) # label anyone 59 and younger at interview who responds no to PFQ049, PFQ057, PFQ059 as "No difficulty" inx_good_fn_C <- which(Covariate_C$PFQ049 == 2 & Covariate_C$PFQ057 == 2 & Covariate_C$PFQ059 == 2 & Covariate_C$RIDAGEYR <= 59) inx_good_fn_D <- which(Covariate_D$PFQ049 == 2 & Covariate_D$PFQ057 == 2 & Covariate_D$PFQ059 == 2 & Covariate_D$RIDAGEYR <= 59) Covariate_C$Difficulty_Walking[inx_good_fn_C] <- "No Difficulty" Covariate_D$Difficulty_Walking[inx_good_fn_D] <- "No Difficulty" Covariate_C$Difficulty_Stairs[inx_good_fn_C] <- "No Difficulty" Covariate_D$Difficulty_Stairs[inx_good_fn_D] <- "No Difficulty" Covariate_C$MobilityProblem <- factor(as.numeric(Covariate_C$Difficulty_Stairs == "Any Difficulty" | Covariate_C$Difficulty_Walking == "Any Difficulty"), levels=c(0,1), labels=c("No Difficulty","Any Difficulty"),ordered=FALSE) Covariate_D$MobilityProblem <- factor(as.numeric(Covariate_D$Difficulty_Stairs == "Any Difficulty" | Covariate_D$Difficulty_Walking == "Any Difficulty"), levels=c(0,1), labels=c("No Difficulty","Any Difficulty"),ordered=FALSE)
Saving the demographic, lifestyle, and comorbidity data
Finally, we save the columns of interest from these data files
vars_inc <- c("SEQN","SDDSRVYR", "SDMVPSU","SDMVSTRA","WTINT2YR","WTMEC2YR", "RIDAGEMN","RIDAGEEX","RIDAGEYR","BMI","BMI_cat","Race","Gender", "Diabetes","CHF","CHD","Cancer","Stroke", "EducationAdult","MobilityProblem", "DrinkStatus","DrinksPerWeek","SmokeCigs") Covariate_C <- Covariate_C[,vars_inc] Covariate_D <- Covariate_D[,vars_inc] save(Covariate_C, file="Covariate_C.rda", compress="xz") save(Covariate_D, file="Covariate_D.rda", compress="xz")
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.