The Business Survey Methods Division of Statistics Canada is looking for an individual for a paid internship to work on a project to investigate calibration in the context of Business surveys. The candidate would preferably have or be working towards obtaining a graduate level degree and have knowledge of R or SAS and calibration estimation theory in the context of survey sampling. More details are available in the attached document. We are asking the network of CANSSI institutions to help identify such individuals. If you are aware of appropriate candidates, could you please share this information with them? Candidates have until December 15th, 2016 to express their interest. The length of the internship is four months, with a possibility of extension, and will start in the winter of 2017 (ideally in January or February, 2017) according to the availability of the candidate and will take place at the Statistics Canada offices in Ottawa.
If you or they have any questions, please do not hesitate to contact Marie-Claude Duval (Marie-Claude.Duval@canada.ca).
Calibration project for a student at Statistics Canada, Ottawa
Beginning of 2017 – for 4 months*
The calibration working group in Statistics Canada’s Business Survey Methods Division (BSMD) is seeking a student to conduct calibration studies using business survey data and administrative data.
The working group has raised various issues in the results of the calibration method and the constraints used by some economic surveys in the Integrated Business Statistics Program (IBSP). These include significant variations in weight in some cases after calibration, significant variations in estimates and/or variances compared with a previous survey cycle, no solution usually due to too many constraints, and a lack of or limited diagnostics to assess and validate the data and calibration method.
The student, who will be supervised by a methodologist with experience in the field of calibration, will use business survey data and auxiliary data to make diagnoses to identify and understand problems encountered and then report on the desired or necessary conditions for calibration to be effective. For example, conditions could include the minimum sample size needed for calibration, the minimum correlation needed between survey variables and auxiliary variables, the choice of calibration groups, the choice of auxiliary variables, validation of the model and the treatment of outliers, out-of-scope units, and units incorrectly classified in the calibration process.
If time allows or if the project is renewed, the student could be called upon to propose other, more effective estimation approaches than calibration in some situations. The student could also collaborate in the development of an analysis tool (or specifications) to evaluate data, validate the model, and correctly specify the parameters to be used in calibration in an effective manner.
The results of this project should allow the working group to provide guidelines that will help methodologists to make informed and effective decisions in justifying whether or not to use calibration in their economic surveys, particularly surveys that are or will be integrated in the IBSP.
Software: R or SAS
Technical knowledge: Calibration estimation theory.
Four months with a possibility of extention. The student could begin in January or February 2017, depending on the student’s availability.