sachsmc/cosinor

R package for estimation and prediction of the cosinor model for periodic data

cosinor's Issues

Add plot function

cosinor.lm does not work with tidy dataframes.

I noticed that cosinor does not work with tidy dataframes:

dat <- readr::read_csv(data_location)
class(dat) # "tbl_df"     "tbl"        "data.frame"

mod <- cosinor.lm(y ~ time(time), period = 1440, data = dat)

This gives an error:

Error in model.frame.default(formula = newformula, data = data, na.action = na.action,  : 
invalid type (list) for variable 'rrr'

The error occurs in this line:

data$rrr <- cos(2 * pi * data[, timevar]/period) # from body(cosinor.lm)

Running just this line for my data set gives the error message:

Error: Unsupported index type: function

I don't know if this should be the responsibility of the user or not... Please ignore this if that is the case!

Implement mixed effects cosinor model

Using lmer or lme, See R code in the appendix of Albert & Hunsberger, Biometrics (2005).

Out < −nlme(conc ∼ (A + exp(B) ∗
(C ∗ cos(2 ∗ pi ∗ (time + D + exp(E)/(1 + exp(E))))
+ F ∗ cos(4 ∗ pi ∗ (time + G + exp(E)/(1 + exp(E)))))), 
data = dataset, 
fixed = list(A ∼ Group, B + E ∼ −1 + Group, C + D + F + G ∼ 1), random=A+B+E∼Group|ID,
start = c(1.8, .2, .1, .2, .5, .1, .2, .1),
control = list(maxIter = 100, returnObject = T))

Write vignette

Extension of ggplot.cosinor.lm

Dear Michael,

I am hoping to use the function ggplot.cosinor.lm to plot multiple cosinors. I have looked into the source code of this function to see if I can extend the accepted covariates beyond 0 and 1. However, I am quite new to cosinor analysis and very new to R!

Looking at the source code, I believe what I would like to extend is this section here:

`for(j in covars){
newdata[,j] <- 0
}
if(!is.null(x_str)){

    for(d in x_str){
        
        tdat <- newdata
        tdat[,d] <- 1
        newdata <- rbind(newdata, tdat)
        
    }
    newdata$levels <- ""
    for(d in x_str){
        
        newdata$levels <- paste(newdata$levels, paste(d, "=", newdata[,d]))
        
    }
    
    
}`

I thought I could change it such that tdat[,d] <- d and this would mean that d in (x_str) would take all values in x_str from 0 to 20 (or however many). But this returned an error: 'variable x_str was fitted with type numeric but type character was supplied'. So I think I messed something up.

Rather than having a covariate of values 0 or 1 (two 'groups') like in the vitamind set, I would like to extend this to more than two groups. Is this possible, or am I missing some fundamental principle?

Thanks in advance, and many thanks for writing this package in the first place :)
Victoria

Fix S3 methods for CRAN submission