Les différences entre les sondages, une question de méthodes? Differing estimates, a question of methods?

Bonjour,

Voici le lien au texte publié aujourd'hui dans La Presse Plus aujourd'hui.

Les tableaux suivants montrent les données sur lesquelles je me suis appuyée pour ce texte. Il est intéressant de regarder les médianes qui donnent normalement une bonne image de la situation.

For English readers, 

The first two rows of the table show the impact of mode of administration on the estimates for the main parties. In short, there is no major difference for Canada as a whole but there are indeed differences according to methods for Ontario and Quebec, the two largest provinces. IVR in Quebec for example has significant differences with classic telephone polls for the four major parties. The analyses controlled for change over time and for other methodological features like proportion of non-disclosers, sample size and number of days in the field.

The two next rows show the minimum and maximum estimates as well as the median for each party over the last week (7 polls). It shows that these differences seem to be somewhat larger than expected, particularly for the Liberal Party of Canada in Ontario and for all the parties in Quebec -- a range between 8 and close to 14 points (BQ). The median normally gives a more reliable portrait of the situation than any individual poll.