Skip to main content
SAGE
Search form
  • 00:07

    SPEAKER 1: In this chapter, we turnto a more traditional approach to uncertainty, which iscalled null hypothesis testing.You may have heard of p-values or significancetesting, both of which are part of the same thing.

  • 00:21

    SPEAKER 2: So what is null hypothesis testing?Null hypothesis testing calculates the probabilitythat the sample effect size that we actuallyhave could have come from a completelyhypothetical population with an effect size of zero.Typically, the null hypothesis canbe rejected when p, the probability value,is less than 0.05.

  • 00:49

    SPEAKER 2 [continued]: The value of 0.05 is called alpha.When p is greater than 0.05, we have failedto reject the null hypothesis.

  • 00:60

    SPEAKER 1: When we reach a conclusion with null hypothesistesting, we may have made one of two different areas becauseof sampling error.Type I error, a false positive whenwe reject the null hypothesis even though thereis no effect in the population.Type II error, a false negative whenwe fail to reject the null hypothesis even though thereis an effect in the population.

  • 01:24

    SPEAKER 1 [continued]: Neither type 1 or type II errors can be identified.It is not possible to know whether we have made an erroror not.

  • 01:32

    SPEAKER 2: Before we begin a piece of researchand looking ahead, we could be about to make either a typeI or a type II error.We can expect that there is a 5% probability that wewill make a type I error.We cannot know what the probability will be that wewill make a type II error.

  • 01:53

    SPEAKER 2 [continued]: At this stage, we could be about to make either.After we have done the research and looking back,we could have made an error.If we have a significant result, then usually the p-valuewe get is an indication of how likely it isthat we've made a type I error.If we have failed to reject the null hypothesis,then we cannot know how likely it is that we have made a typeII error.

  • 02:19

    SPEAKER 2 [continued]: This time we can only have made either a type I error or a typeII error, never both.

  • 02:26

    SPEAKER 1: The first point we careabout is the question that null hypothesis testing answers.The question is about the existenceof a relationship between two or more variables, notthe strength.We use the word effect for this.We want to ask the question, can webe fairly sure that we can accept that there is an effect?

  • 02:47

    SPEAKER 1 [continued]: In null hypothesis testing, we actually ask this question.Can we be fairly sure that we can reject the hypothesisthat there isn't an effect?It is the opposite of the hypothesiswe are really interested in, which isthat there is a relationship.In statistical terms, this is calledthe alternative hypothesis.

  • 03:08

    SPEAKER 2: The answer is still uncertaineven though it appears definite.Up until this point, we have beentalking about the uncertainty that comes with a sample.

  • 03:18

    SPEAKER 1: Suddenly, we are makinga black or white decision.Think of it as a method that exists because researcherscannot tolerate the idea of uncertainty.Therefore an arbitrary rule was madeto provide a definite outcome.The uncertainty is still there.We are uncertain whether we may have made a type I or type IIerror because we cannot know, but the rule allows us to hidethe uncertainty.

  • 03:43

    SPEAKER 2: Null hypothesis testing leads to two outcomes--reject or fail to reject the null hypothesis.Reject the null hypothesis, we are making an inference.Fail to reject the null hypothesis,we are not making an inference.We have failed to infer anything.

  • 04:04

    SPEAKER 1: For more information, see study.sagepub.com/statisticsforpsychology.

Video Info

Publisher: SAGE Publications Ltd.

Publication Year: 2019

Video Type:Tutorial

Methods: Null hypothesis, Type I errors, Type II errors

Keywords: hypothesis testing; null hypothesis; uncertainty and error

Segment Info

Segment Num.: 1

Persons Discussed:

Events Discussed:

Keywords:

Abstract

Null hypothesis testing calculates the probability that the sample effect size could have come from a completely hypothetical population with an effect size of zero. The two different possible error types are discussed.

Looks like you do not have access to this content.

Null Hypothesis Testing

Null hypothesis testing calculates the probability that the sample effect size could have come from a completely hypothetical population with an effect size of zero. The two different possible error types are discussed.

Copy and paste the following HTML into your website