Statistics Made Easy 2: Hypothesis Testing


After you have read the last post on Statistics Made Easy, it should have helped in some way in understanding your classes in statistics.  Note that these Statistics Made Easy posts are to supplement your knowledge in statistics and to pinpoint some of the main important things in statistics; it is in no way going to replace your textbooks in statistics for psychology.

For this post, it will cover the steps of hypothesis testing.  It is one thing that will definitely be mentioned in all statistics classes, and forms the basis for all inferential statistics.  I will cover inferential statistics in the next few Statistics Made Easy post.
One of the direct applications of hypothesis testing in psychology is the testing of the psychological theories when you start applying them to people or clients, and it is your job to test them out (non-statistically).

Steps:
1. Formulate your hypothesis/es
Come up with a null hypothesis and alternative hypothesis (Null = 0 = no difference/relationship/ etc.)

2. Select the appropriate test statistic and level of significance
Based on the type of conditions you are investigating (differences between 2 groups or more than 2 groups/relationship/regression), you can choose the respective tests. 
Level of significance (alpha) is usually chosen between .01, .05, and .10. The alpha of .05 is usually chosen in psychology research.

3. Compute the critical (cut-off) value
This critical value is usually derived from the respective tables (usually found in your textbooks, if you are doing this manually).  The critical value for the test-statistic is determined by the level of significance.  For example, with alpha of .05, you will have to use the critical z-value of 1.645.  The critical value is the value that divides the non-reject region from the reject region.  That meant that with an alpha of .05, you will have to reject the null hypothesis if your computed test-statistic (z-/t-value) is more/higher than the critical value of 1.645.

4. Compute the appropriate test-statistic and compare the computed test statistic with critical value
Here is where you have to calculate the test-statistic, and compare it to the critical value above, which will then make full sense in the next step that follows.

5. Interpret the decision
If the test-statistic is more than the critical value, then you will have to reject the null hypothesis, and accept the alternative hypothesis.  On the other hand, if the test-statistic is less/smaller than the critical value, then you will have failed to reject the null hypothesis, and hence reject the alternative hypothesis.
Advice: In all cases, it is quite impossible to accept the null hypothesis, unless there is truly no difference at all.

Here are the steps of hypothesis testing, which usually happens in quantitative research, although it might not be so explicitly stated when you reach more complex inferential statistics.  In life and work, we often do this as well, testing the educated guesses/hypotheses we have across people or situations and trying to figure out if we are able to believe something or if it is convincing enough.  In life, it's often done without the statistics, but the steps still apply pretty much in the same flow. 

No comments :

Post a Comment