pdbailey0
commented
1 year ago I would approach this by making a new factor variable that has the ESCS bin. The rest of this shows code for this
You must already have done this, but for completeness, I'm showing downloading the data
downloadPISA("~/", years=2015)then I read in data for one country and calculate the percentile. Here I do three levels.
ita <- readPISA("~/PISA/2015", countries="ita")
percentile(data=ita, "escs", c(33, 67))This gives my percentiles of 33rd at -0.5058 and 67th at 0.3907
then I create the bin as a character
# make a low bin
ita$escsBin <- ifelse(ita$escs < -0.5058, "low", "other")
# make mid bin, keeping the low bin
ita$escsBin <- ifelse(ita$escs >= -0.5058 & ita$escs < 0.3907, "mid", ita$escsBin)
# make the high bin, keeping the other two bins
ita$escsBin <- ifelse(ita$escs > 0.3907, "high", ita$escsBin)I then made some summaries to confirm this is correct (not shown).
then I made the escsbinf which is a factor. This allows me both to use achievementLevels and puts the bins in order.
# now make a factor variable
ita$escsbinf <- factor(ita$escsBin, c("low", "mid", "high"))
# use that for the achievementLevel
achievementLevels("read", "escsbinf", data=ita)I hope this answers your question.
ainhoavega
If I understand it correctly, achievementLevels calculates achievementLevels by groups (e.g. by sex male/female). However, the “grouping” variable that I’d like to use is not categorical, but continuous (it’s the ESCS index in PISA).
Therefore, first I have to “transform” it from continuous to categorical, using the percentile function (I’d create 5 groups based on quintiles) and then I’d be able to pass it on and use it as a grouping variable in achievementLevels.
This would allow me to calculate, basically, the % of students who achieve a specific level and fall in the lower quintile of the ESCS, vs. the % of students who achieve that same specific PISA level but fall in the upper quintile of the ESCS.
The workaround I’m using so far is subsetting data based on the quintiles calculated using the percentile function, but it’s not very elegant/efficient. So I was wondering whether what I was thinking is possible or not using the current achievementLevels and percentile functions. Thanks!