[Paired T-Test] Assumptions & Violations

Dason

I'm probably not helping here but... that psychologist is a jerk and needs to learn better experimental design.

Edit: Might as well add something constructive - I think on part vi they actually want you to report the confidence interval and not just state some facts about it...

Karabiner

TS Contributor
Name the dependent variable and its scale of measurement?
Number of assignments not turned in by students after treatment
You forgto to name the scale.
Do positive suggestions versus negative feedback affect the number of assignments completed by a sample of eighth grade students?
I guess you cannot answer that by using such a flawed research design. But anyway.
1c) Describe why the psychologist selected this statistical test?
To determine whether positive rather than negative feedback affect future completion of assignments.
Isn't the question about why he used the dependent samples t-test instead of, say, a chi-square or an independent samples test or something else.
List the statistical assumptions for this design.
There will be a correlation between positive and negative feedback and missing assignments. The variance from each treatment will be equal. Students will have less missing assignments after receiving positive feedback rather than negative feedback.
You're mixing statistical assumptions for performing the t-test with research hypotheses.
Rather, stick to the assumptions of the dependent samples t-test. And read carefully.
Maybe there's a difference between independent and dependent samples t-test, e.g.
regarding equality of variance assumption. And your use of "correlation" here might
distract you from more important topics.
Discuss the strengths and weaknesses (e.g., violations of assumptions) of this study's design.
Amongst other, think about the normality assumption and whether violation of this
assumption can be ruled out.
Students missed assignments less after receiving negative feedback (Mean = 7.25, Std. error .479) rather than positive feedback (Mean = 3.50, Std. error = .645, Correlation = .135, Paired diff. Mean = 3.750).
How can 7.3 missed assignments be less than 3.5 missed assignments?
v) Provide an effect size measure, which may require part-hand calculation, and include your interpretation in your conclusion (above).
There was no significant change in students feedback from the psychologist.
Search for "Cohen's d" for dependent measures.
Provide confidence interval results for t-tests and post hoc tests from SPSS output only.
There's no post hoc test here. And the 95% confidence interval can be found in the output.

Regards

K.

noetsi

No cake for spunky
This reminds me of what I have long thought an anomaly in methods. Its common to learn methods and not learn the experimental design that allows you to understand how to set up the experiment - which is commonly the most important part. So you end up with invalid results.

I took regression and ANOVA over a year ago in my graduate program (and learned the former to some extent over two decades ago). Only next semmester will I finally take experimental design. It should be the other way around.

Karabiner

TS Contributor
This reminds me of what I have long thought an anomaly in methods. Its common to learn methods and not learn the experimental design that allows you to understand how to set up the experiment - which is commonly the most important part.
That might explain why so many questions here in this forum are of the kind
"I have performed an ANOVA. Now tell me something about the interpretation of
technical detail XY" , not "I am studying XY, I set up a study with design XY and
measured XYZ, my sample size is NN, and now I want to answer questiion XY..."

Regards

K.

.

Karabiner

TS Contributor
It lacked 10 participants & was not normal, so it is a relatively useless study?
I don't know where the figure "10" comes from. And I do not know exactly what you mean by "the design
was not normal" and/or where you know it from that something was not normal. And a study is not useless
only because a certain statistical technique cannot be used to analyse its data.

Regards

K.