Random Sampling
I’m not going to say who, but one MSM journalist thinks my recent survey is not ’scientific’. This person did not justify the comment by explaining why, but I would like to address this misconception.
My survey does not meet the standard OLS assumtion of random sampling because I do not know why or who failed to take the survey. But this problem is encountered even by pollsters when using a random number dialer. You can dial people up at random, but if they are sick or not interested or have their reasons which they do not disclose, you know very little about why people did not take the survey.
Social Sciences rarely attain the level of ’scientificness’ that a controlled experiment does. If this is where the criticism comes from, then it is a criticism against all Social Sciences, not just my survey. Therefore, one would have to disregard just about every poll ever taken.
People do engage in misrepresentation and often over-state their preferences in polls and surveys, especially if they feel the need to ‘win’.
What is scientific, however, is recognizing sources of bias and how your sample compares to the population. Bloggers and blog readers are the population of analysis here, and they are not necessarily representative of the Canadian population at large. I am looking at a subset of the Canadian population and defining that as my population. Being scientific (in the Social Science) means that your study recognizes where its biases are. My survey of bloggers found a strong preference for Canadian politics, but that does not mean Canadians in general want to vote that way – but bloggers and their readers might. See the difference?
More to come.
