Output centiles from growth reference for a given measurement

[tamuri]

• summary: generates table of numbers to allow user to plot growth chart
• user-defined centiles set by number or space between centiles
  • equal and unequal centile space
• for each sex
• select age unit and frequency
  • by default, use ages defined in the growth reference
• select growth measurement (height, weight etc) to output
• min and max range is determined by the growth reference

[tamuri]

Centiles form from LMSgrowth:

[tamuri]

Using sitar::LMS2z

## construct table of boys weight centiles by age for WHO standard
data(who06)
zs <- -4:4*2/3 # z-scores for centiles
ages <- 0:12/4 # 3-month ages
LMS2z(ages, as.matrix(zs), sex = 'm', measure = 'wt', ref = who06, toz = FALSE)

[statist7]

Hi Asif,

I’m not sure of the question… Are you pointing out that this code already does #8 ?

Best wishes, Tim

[tamuri]

Sorry, Tim, it wasn't a question! I was sharing the code with @giordano who is currently implementing this feature.

[giordano]

Other things to implement in the centile tab (pull request #26):

• [ ] figure out how to properly round the numbers in the table. See Tim's paper
• [x] related to the point above: store the number of decimal places and significant digits as global variables? (PR #31)
• [x] plotting plot under each table. Note: we may need to plot more datapoints than what the user asks for the table, to smooth out the curves (maybe do two different calls to LMS2z, one for the table and the other for the plot). Look to other plotting libraries (plotly?) (PR #31)
• [ ] for plotting we need to somehow deal with disjunctions, e.g. in who06 data set. Record the presence of disjunction and the year where it happens in a global variable?
• [x] consider using the R package DT (DataTable) for rendering the table in a nicer way (PR #31)
• [ ] We need to rethink the role of the centile table now we have the plot available. Could the table display be toggled with a button, perhaps next to the sexes?

Update: DT makes handling the rounding much easier, you can control how to round each column

[tamuri]

From Tim (https://github.com/UCL/LMSgrowth2/pull/43#issuecomment-602607350):

Some curves have duplicated ages, e.g. UK-WHO height and BMI at age 2 and 4 years, indicating that two distinct references meet at that age. Ideally they should plot both values at the duplicated age, so the line segment is vertical. But in practice just one of the two values is stored and plotted. Not sure how to handle this.

[tamuri]

From Tim (https://github.com/UCL/LMSgrowth2/pull/43#issuecomment-602607350):

The age range for each reference is displayed. However this is not necessarily the age range for the current measurement, so there are missing data at the start of he range. Would it make sense for the age range to apply to the currently selected measurement?