Statistics HELP!?

Question

Help - I have a research methods exam tomorrow. As a hypothetical experiment I have written 'Individuals will recall more words in a memory test in a naturally lit room than in an artificially lit room.' The quality of the experiment isn't the issue - the issue is what test I would use to analyse the date? I thought a 'related t-test'????? Please can someone help, tried to ring my lecturer today to clarify but no answer. Exam is in the morning and I need sleep lol. Thank you everyone your help is appreciated. xxx

Sorry forgot to write that i'm using a repeated measures design, i think that is within groups

Answer 1

It depends ... you need to be more specific. The t test is a test of differences between two means. The independent t test would be used if the two means come from independent groups (i.e., it is a between-subjects design). The related t test is for means obtained by repeated testing on the same people (i.e., a within-subjects design). The experiment you describe could be run with either design (though if you use the same test items for each condition it's probably safer to use an independent measures design to avoid ceiling effects when you test people twice, although the related design is statistically more powerful). The precise nature of the dependent variable also influences the likely choice of test - in a classic word list experiment most people tend to use t tests, but there are some memory measures that might not be suitable for a t test (because the assumptions of the test concerning normality of residuals etc. might not be reasonable).

Note that the choice of design (independent or related groups) influences important aspects of the study ... notably the most important experimental controls: random assignment to conditions for independent measures and control of order effects for related designs (e.g., via counterbalancing).

Answer 2

It depends whether you will be using an independent design or a repeated measures design. With a memory experiment, you will probably be using an independent design (i.e. one group does condition A and another group does condition B). If this is the case, then a related t-test would not be suitable.
You also need to know what kind of data you are collecting (e.g. ratio/interval data) and consider whether it would be appropriate for a parametric test or a non-parametric test. Given the nature of your experiment it would probably be the latter.
Without these bits of information it is difficult to make the right decision.
You may have a flow-chart to use so you can get the right one; you could explore Mann-Whitney U test.

Answer 3

You would use a within (also called "paired" or "repeated measures" or "correlated") groups t-test. This test will compare the subjects' data under the artificial and naturally lit conditions.

If the two groups do NOT have the same people in them, then you would do a between (also called "independent") groups t-test.

Good luck!

Answer 4

If my notes weren't out in my car, I'd help you out here right quick...but as it is...they are across the parking lot and I won't see them for another couple of hours (I had stats last semester and now doing behavior/learning psych...our first test in a couple of weeks is on methodology).

Answer 5

Sounds like whoever set up the experiment is a little loopy|






---