Weighting methods in surveys / Painotusmenetelmät surveyssa
Weighting methods in surveys, spring 2016 / Painotusmenetelmät surveyssa, kevät 2016
Lecturer Luennoitsija
Basic targets Tavoitteista
The course covers 6 times, 3 hours each, on Thursdays 16-19. The weeks and and the computer class are below. Each event includes both lectures and trainings. The weighting is required if wished to generalize survey output to a target population. If you just calculate without any generalization target, it is not needed. Naturally, weights are necessary in sample surveys but also in the case of unit nonresponse in complete population studies (thus if the sample is 100%). This last point is not always recognized and omitted, and hence the estimates are biased.
The course content includes:
- Importance of weighting in surveys, in sample surveys in particular
- Generalization
- Specific cases when weights are used
- Before weighting: inclusion probabilities with most common sampling designs (srs, stratified srs, multi-stage)
- Creation the different weights: gross sample design weights, basic sampling weights for net sample, post-stratification weights, response propensity weights, calibration weights, joint propensity and calibration weights
- Two-level weights from single-level weights, panel weights and other special weights
Both theory and practice are integrated with empirical micro data. Since many things are easiest to do by SAS, not SPSS, we use SAS but SAS codes and other important things are given for trainings. Some basics on SAS is good to know in advance. It is possible also using SAS: Open SAS at least and look at the windows there. There you see windows Editor and Log, and as soos as you have made something, you will get the result output. The SAS codes are to be copied and further handled in Editor, and Log gives you the diagnostics of your Run (push run to do something). All documentation is available via Help. Please look at some of these Help options (up in the screen). All details are in SAS Help and Documentation. Here you can find even Learning to Use SAS. I thus hope that you have opened SAS and tried to do something before the course start. In the first event, we will start to use SAS with simple examples, and go to learn more. In the very late of course, we will use SAS macro Calmar2. This is a basic reason to use SAS. This French macro is not available in other software.
In Finnish, basic things too:
Tämä on joiltakin osin samaa kuin surveymetoodiikan kurssi mutta sillä ei oikein kunnolla kyetä käymään läpi esimerkiksi sisältymistodennäköisyyksien käytäntöä. Nyt siis, jos kyseessä on otos kuten yleensä, tarvitaan näiden muodostaminen jotta tuloksia voidaan yleistää eli estimoida halutun tavoiteperusjoukon tasolle. Yleistämisen ensimmäinen edellytys on että otos- ja muut tarvittavat painot olemassa. Kurssilla käytetään oikeanlaista dataa eli paria dataa ja mm. noiden todennäköisyyksien merkityksen luulisi tulevan selväksi. Mutta tästä jatketaan koska kaikissa aineistoissa syntyy eriliaisia häiriöitä. Tällä kurssilla päähäiriöitä eli harhan aiheuttajia ovat vastauskato ja peittävyyden ongelmat. Näihin vastataan nyt myös painotuksella mikä jatkokäytössä hyvien painojen tultua tehdyiksi, on melko helppo verrattuna toiseen menetelmään eli imputointiin.
The number of the credits is five if the minimum number of tasks are done. If more work is done, more credits are possible, until seven.
Welcome
We can use both English and Finnish in lectures and trainings.
Schedule for Lectures and Trainings
Six Thursdays since 28 January until 3 March, 3 academic hours beginning at 16:15, in a computer class. The place will obviously be Aleksandria, K131.
Exam
The course will be done doing the basic trainings and reporting about them. It is possible to arrange a small-scale exam event (about one hour) about those things that are not covered in the training.
We will see this during the course.
Materials
It is good to read the attached paper that helps in understanding the starting point of weighting. I will thus give some sampling data files that are used for weighting.
The entire course material is below. In addition, SAS codes are sent to registered participants by email. Enjoy reading and understanding.
Some references on weighting
Register to the course / Ilmoittaudu kurssille
Unohditko ilmoittautua? Laita viestiä osoitteeseen tilasto-info[at]helsinki.fi