Section D: Inferential Testing Flashcards
What are the THREE levels of measurement? (think data types!)
-NOMINAL
-ORDINAL
-INTERVAL
What is NOMINAL data?
-Data that falls into CATERGORIES OR GROUPS.
-The most BASIC level of data –> Categorical with a total number of values in each group.
(e.g. favourite flavour of ice cream of the members of UVI).
OR
(e.g. a horse in a race may run very fast, fast, average or slow. This does not tell us the actual race time of each horse).
Interval and ratio data does show the differences between each value and there is a scale used where the intervals between each value are equal. The difference between interval and ratio is that ratio data will have an absolute zero. Examples include height in cm, time in minutes and weight in pounds because they all start with zero.
What is ORDINAL data?
-RANKED data or data on a made up scale.
-Ordinal data is more detailed –> In a race we will know each person’s position (1st, 2nd, 3rd etc.)
(e.g. how happy students are on a scale of 1-10 - highest to lowest or lowest to highest).
(e.g.) in the race we will know each horse’s position (1st, 2nd, 3rd etc). However, it does not tell us the difference between each horse.
What is INTERVAL data?
Data that tells you specifically where someone is on an OFFICIAL AND STANDARDISED SCALE.
-Data does show the differences between each value and there is a scale used where the intervals between each value are equal.
(Actual units of measurement on standardised scales - centimetres, heart rate, blood pressure, temperature in degrees celsius, time in seconds).
Choosing an appropriate statistical test: What are the 3Ds you must consider?
- Is the research a test of difference or relationship?
(A test of difference involves comparing the data of two or more groups to see if there is a difference/ A test of relationship is looking to see if one variable is related to another.) - What design has been used?
(If a test of difference you will need to consider which design was used - repeated measures, matched pairs, or independent group/ if a correlation has been performed then each participant provides both measures, so you need not differentiate). - What level of data (measurement has been used)?
(Is it nominal, ordinal or interval?).
When is the sign test used?
-A test of DIFFERENCE.
-Nominal (categoric) data.
-Related Design (Repeated Measures/Matched Pairs).
HOW is the sign test calculated? - OUTLINE THE SPECIFIC STEPS
- Working out the difference between the two sets of data –> Involves SUBTRACTING one set of data from the other (before - after). PUT A + OR - BY EACH PPT.
- ADD UP THE TOTAL NUMBER OF +/- (WHERE THERE IS NO DIFFERENCE - 0 - THE DATA CAN BE IGNORED/ PPTS DELETED IF THERE IS NO DIFFERENCE BEFORE AND AFTER.
- The LESS FREQUENT SIGN (OF + AND -) = OUR CALCULATED VALUE (S = ….)
To work out the N value (the amount of ppts) –> We take number of ppts BUT WE discount all of those with a difference of 0 –> They are deleted (no of participants - ones where the difference is O).
- Compare your CALCULATED value to your CRITICAL VALUE (found in critical values table) –> To find right value - we ask ourselves is the test ONE TAILED or TWO TAILED? What if the significance level being used - unless told otherwise the significance level is 0.05 AND how many participants there are - then find this in value table.
(This is given to you in exam! but just to understand) - In order for results to be SIGNIFICANT, what must the calculated value of ‘S’ be?
-The calculated value of ‘S’ must be EQUAL TO or LESS THAN the critical value in order to be significant.
What is the magic paragraph - determining whether something is statistically significant?
The calculated value of S = …. IS/IS NOT significant at the 0.05 level for a TWO TAILED HYPOTHESIS where N = …. This is because the CRITICAL VALUE IS ….. Therefore, as the calculated observed value is MORE THAN/ LESS THAN the critical value, results ARE/ ARE NOT SIGNIFICANT.
What does PROBABILITY refer to in psychology?
Probability refers to THE LIKELIHOOD OF AN EVENT OCCURING. It can be expressed as a number (0.5) or a percentage (50%).
What are inferential statistical tests necessary for?
Inferential statistical tests are necessary to determine whether our results are SIGNIFICANT OR SIMPLY DUE TO CHANCE.
-This will show which hypothesis to ACCEPT and which to REJECT.
-Statistical tests allow psychologists to work out the PROBABILITY that their results could have OCCURED by chance, and in general psychologists use a probability level of p<0.05 = (5%) –> The value indicates the PROBABILITY THAT THE NULL HYPOTHESIS IS TRUE.
What is the definition of p<0.05?
The likelihood of the data (in terms of difference or relationship found) being DUE TO A RANDOM CHANCE IS LESS THAN OR EQUAL TO 5%).
-There is a 5% CHANCE THAT THE NULL HYPOTHESIS IS TRUE.
If psychologists used a 0.01 level of probability, what % likelihood would there be that results were due to random chance?
1% LIKELIHOOD THAT RESULTS ARE DUE TO RANDOM CHANCE.
What are the three different significant levels psychologists tend to use?
p<0.10 or 10% - RARELY USED (this is too relaxed).
p<0.05 or 5% - THIS IS MOST COMMON IN PSYCHOLOGY RESEARCH.
p<0.01 or 1% - THIS IS THE MOST STRINGENT –> OFTEN USED IN MEDICAL OR SAFETY CRITICAL SITUATIONS.
What is a TYPE ONE ERROR + How does it occur?
A TYPE I ERROR = A FALSE POSITIVE.
-A type I error occurs when a researcher CLAIMS SUPPORT FOR THE RESEARCH HYPOTHESIS WITH A SIGNIFICANT RESULT WHEN THE RESULTS WERE CAUSED BY RANDOM VARIABLES.
-The alternative hypothesis is ACCEPTED but results are NOT REALLY SIGNIFICANT.
-This occurs when the level of significance is TOO LENIENT - Sometimes called an error of optimism/ not cautious enough.
(e.g. You conclude that the drug intervention improved symptoms when it actually didn’t –> These improvements could have arisen from other random factors or measurement errors).
The boy who cried wolf –> FIRST everyone believed there was a wolf when there WAS NOT (Type 1).
