Page 196 - Marketing the Basics 2nd

188 Marketing: the Basics
used. If the number of observations is over 30, the appropriate test is called a z-test. Other types of distributions include exponential (used to model the growth rate of bacteria for example), Poisson (no it does not measure the growth rate of fish, time periods actually), chi-squared and a host of others. For our purposes, we will restrict ourselves to normal distributions.
Once the appropriate test is chosen, the data is normalized. Now don’t get confused between normal distribution and normalizing the data. Though the data can be modelled by a normal distribution they must first be transformed into a data set that will fit under the bell curve. The mathematics proving that the quality of the data is unaffected is well beyond the scope of this book, so take it for granted like the rest of us and assume it’s true. With the data normalized, the researcher can now test if the null hypothesis is true by assigning a score to the data. This is done by subjecting the hypothesis to the most extreme cases possible, and seeing if the score lies under one of the tails of the curve. Statisticians call this area the alpha value. If the score does lie in that range, then you can safely conclude the null hypothesis is incorrect. If it does not, you do not conclude the null hypothesis is correct. Rather you conclude there is insufficient data to reject the null. Once the null has been determined not to be false (but not necessarily true), many more tests are needed to ensure the finding’s reliability and validity. Repeating the test ensures that no errors were made in data collection, that researchers did not knowingly introduce a bias in the results or that other methodological mistakes were made. It takes a lot to convince a statistician something is true!
One final note about quantitative models. Even if a relationship between two variables is finally proven true using statistics, models can be manipulated. As Benjamin Disraeli, a nineteenth-century Prime Minister of Great Britain once remarked, ‘there are lies, damn lies and statistics’. Even today you can still read stories in the news- paper that claims a scientific study proves that one gender is better at logical problems than the other, or worse, one race is smarter than another. Our ice cream driver believes temperature affects ice cream sales, but forgets that if it is raining buckets outside, even if it’s 30 degrees Celsius outside, very few will make the effort to run for their ice cream.