Skip to main content
SAGE
Search form
  • 00:00

    DANIEL LITTLE: In this video, I wantto discuss the goals of statistical testingin a very general sense.In particular, I want to give examples of caseswhere statistics can be useful.I'll talk about what happens if you have a single sample.For instance, if you have a single data pointfrom some hypothesized distribution,one question you can ask is whether that data point came

  • 00:22

    DANIEL LITTLE [continued]: from a specific distribution.A second question would arise if you have more than one datapoint.If you have a whole sample of data,then you can ask whether the entire samplecame from some specific hypothesized distribution.Finally, a third situation would arise if youhave two or more samples.

  • 00:42

    DANIEL LITTLE [continued]: In this last case, the question is, did these samplescome from the same underlying distribution?In addition to considering how we approacheach of these goals, I also want to examinethe meaning of the p-value in each of these cases.For each of these cases, we would use some statistical testto arrive at a p-value, which we would then

  • 01:03

    DANIEL LITTLE [continued]: use to make some inference about our sample of data.So what are some issues that we might encounter that maponto each of these questions?One thing you can do is imagine that you'rea clinical psychologist who is interested in determiningwhether a child that you are treatingis exhibiting antisocial behavior whichdiffers from the normal population

  • 01:24

    DANIEL LITTLE [continued]: of antisocial behaviors.Given the distribution of behavior,you can ask how typical or atypical is that single childto that single point of data.Alternatively, you could imagine that you'rea forensic psychologist who has developed a new line-upprocedure.And you're interested in determiningwhether your new line-up procedure improves

  • 01:46

    DANIEL LITTLE [continued]: the performance of a single group of participants.One question you can ask is whether your new sampleof participants perform significantly betterthan chance.That is, are they doing better than just guessing?Finally, you might imagine that you're a researcher for a drugcompany who has developed a new drug to treat high bloodpressure.

  • 02:06

    DANIEL LITTLE [continued]: And you're interested in determiningwhether the new high blood pressuremedication reduces blood pressure more than some placebodrug.In this case, you could compare the two groups.One who receives the actual medication and onewho receives the placebo.The question you want to know hereis whether or not these groups actually differafter receiving the drug.

  • 02:27

    DANIEL LITTLE [continued]: Our main goal in all of these casesis one of statistical inference.In particular, we want to know howwe can determine from a single observationor from a set of observations whether or notthose observations are somehow weird or unusual.One thing we have to remember is that data are variable,they're noisy.They don't reflect only the process we're interested in,

  • 02:50

    DANIEL LITTLE [continued]: but also random variation.What we're really asking is whether the samplesthat we observe are generated or sampledfrom the same underlying population.In order to understand these concepts,it's important to understand a related concept, whichis the distribution of data, or a distribution of data.

  • 03:12

    DANIEL LITTLE [continued]: A distribution is simply the different valuesor range of different values that make sense as valuesobtained from an experiment.So for instance, in the example that I've shown here,on the horizontal axis I've listedthe height of possible people that we might observe.And the heights range on the lower end from 4.5 feet

  • 03:34

    DANIEL LITTLE [continued]: all the way up to 7.5 feet.What the bars indicate on the vertical axisis something about the probability of observing eachof those particular heights.The probability might be representedas just the frequency or proportion of timesthat we could expect to encounter each of these values

  • 03:57

    DANIEL LITTLE [continued]: if we run the experiment indefinitely.We all have an intuitive sense of statistics.For instance, we would find a person with a height of 7 feetto be very unusual, because we're notused to seeing people quite that tall.But a person with a height of 5-- who is 5foot 7 inches is not really worthnoting at all because they fall right around the average height

  • 04:20

    DANIEL LITTLE [continued]: that we're used to seeing.Each of these concepts is something we'll explain furtherin these videos.Statistics is a way in which we quantify this intuitive feelingof unusualness when we, for instance, seesomeone who is 7 feet tall.

Video Info

Series Name: Statistics for Psychology

Episode: 5

Publisher: University of Melbourne

Publication Year: 2014

Video Type:Tutorial

Methods: Statistical inference, Sampling distribution, P-value

Keywords: mathematical applications

Segment Info

Segment Num.: 1

Persons Discussed:

Events Discussed:

Keywords:

Abstract

In chapter 5 of his series on statistics for psychology, Professor Daniel Little presents a series of hypothetical scenarios for extracting data. He highlights the idea that statistics are often trying to isolate the weird and unusual occurrences out of a set.

Looks like you do not have access to this content.

Goals of Statistical Testing I

In chapter 5 of his series on statistics for psychology, Professor Daniel Little presents a series of hypothetical scenarios for extracting data. He highlights the idea that statistics are often trying to isolate the weird and unusual occurrences out of a set.

Copy and paste the following HTML into your website