# Psychology; Methods. Statistical Inference

First part of the three sections of the "Inferential Statistics" element of the Methods section of the PSYB4 syllabus.

HideShow resource information
• Created by: Gee-wa
• Created on: 29-06-10 19:35

First 443 words of the document:

The Concept of Probability & Levels of significance:
Levels of Significance:
Before an investigation psychologists must decide what level of probability they are prepared
to accept that a set of results occurred by chance.
In psychology usually settle for being 95% sure that results are not due to chance or factors
other than those being studied before accepting the experimental hypothesis.
Therefore accept 5% probability results are due to chance i.e 0.05 level of significance.
Probability:
In some cases a more stringent level of significance is used.
The 0.01 level means the probability of the results being due to chance are less than 1 in
100.
These stricter demands are not usual in psychology but are so in cases like medical research.
0.01 selected only really when seeking to disprove a wellestablished theory.
Hypothesis testing: null and alternative (experimental or research) hypothesis:
Hypothesis:
A hypothesis is a statement predicting the outcome of an investigation.
Hypotheses can predict a difference or an association or correlation.
When a hypothesis predicts a difference of correlation its known as a research hypothesis. If
it is then investigated using experimental methods it becomes an experimental hypothesis.
There are also null hypotheses which predict no significant difference or correlation.
Every study should have a null and experimental hypothesis, the experimental becoming the
alternative hypothesis.
Whatever an outcome both hypotheses cannot be true as they contradict.
Researchers retain the one that is supported and reject the one that is refuted.
This decision is made through inferential testing.
Testing:
If the hypothesis predicts a difference it needs one sort of test, however if it predicts a
correlation it needs another.
The second factor determining the sort of test is the experimental design: Repeated
measures, matched pairs, and independent groups.
Idpt groups Matched/Repeated Relationship
Nominal Chi Sq Sign Chi Sq
Ordinal MannWhitney Wilcoxon signed ranks Spearmans Rank
Order
Interval Unrelatedt Relatedt Pearson's product
moment correlation.
Nominal data: categories
Ordinal data: Data in the form of scores, but not necessarily from a scale with equal
intervals: ranking/rating people.
Interval data: Data in the form of scores taken from standardised scale with equal intervals.
Type I and Type II errors:
Since we can never be 100% sure results are not due to chance we must accept there is the
probability that they could be. This can lead to:
Type 1 error: Wrongful rejection of null hypothesis.
Type 2 error: Wrongly accepting the null hypothesis.

## Other pages in this set

### Page 2

Here's a taster:

The more stringent the level of significance (e.g 1% level) the more likely it is we make a
type 2 error.
The more lenient the level (e.g 10%) the more chance of a type 1 error.
5% is often seen as offering the best balance.
Positive, negative and zero correlation:
Zero correlation: Predicted in a null hypothesis there will be no relationship between
variables.
Twotailed hypothesis: Predicts a difference/correlation but not its direction.
Onetailed hypothesis: Predicts a difference/correlation ant states direction.…read more