PISA 2018 data

[Onur Pehlivan]

Dear all,

I am doing a research on PISA 2018. I've downloaded PISA 2018 data sets from oecd.org/pisa. But, I cannot find students' individual scores (not country's mean score). I need students' individual scores for my research. How and where can I get them?

Thank you very much for the answers.

[Maria Boutchkova]

This is not a Stata question.
I have not used the PISA data but looked through the OECD website, and see that they provide the codebook with all variable names and descriptions.
http://www.oecd.org/pisa/data/2018da..._CODEBOOK.XLSX
Importantly, the student level data is a separate file and is the largest since it is the most dis-aggregated - perhaps you are looking at school or country level files instead of the student one.

[Onur Pehlivan]

For researches who are working on PISA 2018:
Students' scores are in Student Questionnaire Data File as 10 plausible values.

[Onur Pehlivan]

Maria Boutchkova thank you for the answer.

[Maciej Jakubowski]

Hi, you can also download PISA data in Stata format from here: https://www.evidenceinstitute.eu/data/

We are now updating PISATOOLS package to facilitate analysis with PISA data. It should be available in a few days from the SSC repository. Until then, you can use the PISATOOLS package that is available on our website - just specify "cycle(2015)" as an option - it will also work with 2018 data.

Best

Maciej

[Lisa Behm]

Dear Onur Pehlivan , I encounter the exact same difficult. I am required to calculate them myself on the individual level and create a new variable. However, with all special packages (repest, pisatools, pv) I have not managed to calculate and generate an individual test scores variable. How did you proceed in the calculation of test scores on the individual level?

[Ahmad Ghaemi]

Dear Onur Pehlivan , I encounter the exact same difficult. I am required to calculate them myself on the individual level and create a new variable. However, with all special packages (repest, pisatools, pv) I have not managed to calculate and generate an individual test scores variable. How did you proceed in the calculation of test scores on the individual level?

Hi Lisa,

If you read through the PISA manual (I think chapter 5 or 6 is on PV's) you can see that you should not do regressions on individual level. Are you aware of that? What you should do is to first obtain whatever statistic you want for each of the plausible values separately, and then take the mean of them. So for for example, if you want to run a regression with parental income as the explanatory variable, where i is the index for the student:

PV1_i = B1*PARINC_i + Error_i
PV2_i = B1*PARINC_i + Error_i
PV3_i = B1*PARINC_i + Error_i
PV4_i = B1*PARINC_i + Error_i
PV5_i = B1*PARINC_i + Error_i

And then take the mean of the B1's, and report that result. You should NOT first take the mean of the PV's for each individual and then run a regression. So do NOT do the following:

PV_MEAN = PV1 + PV2 + PV3 ... / n
PV_MEAN_i = PARINC_i + Error

... because it will give a biased result.



The pv package in Stata should be able to do this for you. Type "pv help" as a Stata command and then look at the examples with PISA data. I am only familliar with TIMSS/PIRLS so can't help you in detail. But for example, use this and plug in the correct variables instead of pv*math, w_fstuwt, w_fstr, ... etc:

 pv, pv(pv*math) weight(w_fstuwt) brr rw(w_fstr*) fays(0.5): reg @pv stratio propqual [aw=@w]