Page 195 - Marketing the Basics 2nd

Market research 187
start, the researcher must first observe a cause and effect relationship between two events. Let us assume a driver of an ice cream van noticed hordes of children, the children chased after him (driving at a very slow and safe speed of course) for blocks on end when the temperature was above 25 degrees Celsius. Once he stopped the van, a feeding frenzy would take place, as children practically begged him for larger sizes. However, when the temperature was cooler that 25 degrees, only those children who really wanted ice cream would make an effort. Leading the driver to conclude that the temperature is the difference between a bonanza and breaking even.
Is the driver correct? That’s what quantitative models set out to prove one way or another using the principles of statistics, which believes if you observe a phenomenon long enough patterns will emerge. These patterns fall under a certain number of categories called distributions. Choose the right distribution, and you can accurately predict outcomes.
To build a quantitative model, a researcher would first express the relationship between ice cream sales and temperature mathe- matically in the form of a null hypothesis. A null hypothesis can be expressed as an equality, inequality or as the difference between two samples. A null hypothesis is also assumed to be true unless proven otherwise. The reason why statisticians assume the null is true is because it is much easier to disprove truth than prove it. Imagine you had an argument with a friend who claimed that employees are most productive at work on Mondays. You had a sneaky suspicion in the back of your mind that perhaps your friend was stretching the truth. To see if that is indeed the case, you went on the computer, conducted a search on the Internet and discovered in fact your friend did misquote the fact. The reality is that workers tend to be most efficient on Tuesdays and are least effective on Fridays as they impatiently await the weekend. Statisticians love to disprove things as well, except they use mathematics.
Once the null hypothesis is stated, an appropriate test is chosen to determine the probability that the null is indeed true. Choosing the test is not as difficult as it sounds. All you need is the sample size and the type of distribution. Under most cases, the distribution would be the standard bell-shaped curve familiar to university students. If the number of observations is below 30, then a t-test is