Reviewed by Oct 05, 2020| Updated on
A null hypothesis is a kind of hypothesis that is used in statistics to indicate that there is no difference between specific characteristics of a population (or a process that generates data).
A gambler may be interested, for example, in whether a game of chance is fair. If it's fair, then for both players, the estimated earnings per match are 0. If the game is not equal, then the predicted earnings for one player are positive and negative for the other.
The gambler collects earnings data from several repetitions of the game to test whether the game is fair, calculates the average earnings from those results, then checks the null hypothesis that the predicted earnings are no different from zero.
If the average sample data earnings are far enough from zero, then the gambler rejects the null hypothesis and assumes the alternative hypothesis, namely that the estimated earnings per match are different from zero.
If the average sample data earnings are close to zero, then the gambler does not embrace the null hypothesis. Instead, he may conclude that the disparity between the test average and 0 can be explained by chance alone.
The null hypothesis, also called conjecture, implies that every kind of discrepancy is due to chance between the selected characteristics that you see in a collection of data. For example, if the estimated gambling game earnings are truly 0, then any discrepancy between the data average earnings and 0 is due to chance.
Analysts look at rejecting the null hypothesis as it's a clear conclusion. The alternative hypothesis that the findings are "explainable by chance alone", is a poor assumption, because this allows variables other than the chance to be at work.
The statistical hypotheses are evaluated using a method in four phases. The first step is for the analyst to state the two hypotheses in such a way that only one is right. The following step is to come up with a plan of study, detailing how the data will be analysed.
The third step is to execute the strategy and evaluate the sample data in a physical way. The fourth and final step is to evaluate the findings and either deny the null hypothesis or assert that the variations found can only be explained by chance.
The term "null" means that researchers work to nullify it as a widely accepted truth. This does not mean the argument itself is null! (Perhaps the word "nullifiable hypothesis" would be more appropriate so it might cause less confusion).