R-Instat may soon be able to simplify the analysis of IPUMS data in R?

[rdstern]

We have regularly been in touch with the IPUMS team and have mentioned R-Instat to them. I suggest we will soon be ready to return to them and suggest that using R-Instat is an obvious route to using R for their data.

1. They have a package, namely ipumsr on cran. I am not sure we need it! I used it, to supply example datasets. They were xml files that produced an error in our File > Import data. Not such a good start! But that is not highly relevant. They suggest that their real data be accessed as I have done for the 10% sample of the Kenya 2009 census. This is my best example of an SPSS file that is over 1 million records it is 3.8 million by 98 variables and I need to be using the 64 bit version to have enough room!
2. They have an article here about value labels and the difference from factors. However, once imported they suggest we generally use factors. This is all fine with R-Instat as it uses both!
3. They also have an article called Recode and summarise variables from multiple response questions. This is particularly on getting sets of data using dplyr - the way we do with select. That will be part of one way of handling multiple response data. I suggest we complete our select stuff first.
4. Then we should also complete our current Summary and Frequency tables work.
5. And we should add something on multiple response tables. From an initial search this doesn't seem to be well done yet in R. I think the ufs package might have what we need for dichotomies. This where you "tick all that apply", and that gives a set of yes/no or missing variables. (An alternative is "Which of these games have you played in the last 3 months - and you choose each time from the list.) A simple function is in the ufs package. So, I suggest we wait until we have progressed on this.
6. Finally we should do more with the survey package. This is already installed. I will be suggesting we define a srv object - as we have done for survival. And then include it in the Fit Models keyboard.