Which test to use, ANOVA, T-test, other? Urgent

#1
Purpose of Study: Looking at the attitudes of current and former university students about climate change research methods.

Questions:
Do demographic factors have an affect on how participatory ones views are on research design. Do these factors affect how participatory their views are, and how much they agree with given statements.
Does level of exposure to certain types of course content affect how participatory their views are, and how much they agree with given statements.

Nature of Data:
Demographics - age, ethnicity, gender, degree subject, identify as LGBT+?, disability (nominal/ordinal data)
Exposure to __ in degree course, eg sources from climate scientists, sources from intergovernmental organisations (ordinal data)

Dependent variables:
How participatory their views are for research. Measured through multiple choice of preferred actions to take in certain situations. I have ranked the actions from most to least participatory (1-4) and have calculated an average score across the questions for each participant.
How much they agree with statements (ordinal)

My questions:
-With the scores for participatory-ness, would this be counted as ordinal or interval data? The choices are ordered and treated as a 1-4 scale.
-Which stats test or tests should I use to analyse this data. Could it be multiple t-tests, an ANOVA for each question, or would it be a single ANOVA including all IVs? Or would it be something else?
-In the questionnaire, some participants did not fill out all questions such as specific demographic questions. Can these participants still be included in the tests and these questions be left blank or would they need to be removed from some or all of the tests?

Thank you for your help with this!
 
Last edited:

obh

Active Member
#2
1. With the scores for participatory-ness, would this be counted as ordinal or interval data? The choices are ordered and treated as a 1-4 scale.
2. Which stats test or tests should I use to analyse this data. Could it be multiple t-tests, an ANOVA for each question, or would it be a single ANOVA including all IVs? Or would it be something else?
3. In the questionnaire, some participants did not fill out all questions such as specific demographic questions. Can these participants still be included in the tests and these questions be left blank or would they need to be removed from some or all of the tests?
!
If several survey questions measure the same measurement from different angles, you should probably take the average of these survey questions.

1. What did the participants see? 1,2,3,4? "disagree, neutral, agree" ?
2. First ANOVA with 2 categories gets exactly the same results as pooled independent t-test.
What question should the test answer?
3. You may include or exclude the question for different tests, if for example, a test compares DV by demographic, you may exclude the questions without the demographic data, but use these question on other tests. you can also present these questions as "other" on demographic charts
 
#3
Thank you for your help!

1: they were presented with 4 responses they could take to the scenario and had to choose which they would take. The options were ranked from most to least participatory, converted to numbers 1-4, and we found an average score for each participant.

2: I spoke to someone else and they said an ANOVA would be best as I have several IVS in the demographics (age, gender, ethnicity etc) and want to see if any of these affect their average score on the situations questions.
-I am doing 4 tests comparing demographics to situation scores, demographics with agreement with statements, exposure to specific topics to situation scores, and exposure to specific topics to agreement with statements. The demographics and exposure to topics include multiple IVs (each demo characteristic, and exposure to each type of content.

3: that's what we'd assumed, thank you for confirming!

I hope that all makes sense. We think it would be 4 ANOVA tests as they compare the effects of multiple IVs to a single DV each time, but not sure if they would need to be one-way or two-way ANOVAs.
 

obh

Active Member
#4
Hi AC,

ANOVA and linear regression have the same results.
Since you have several IVs you probably should use the regression. How many IVs do you have?

The question is what is your default assumption?
If you know from other researches that all these IVs should influence the DV, doing one way ANOVA is like doing several simple linear regression, each time with a different IV, instead of multiple regression. so other IVS may interfere you one way ANOVA (unless you plan it otherwise)

But doing a multiple regression with many random IVs may find one of them significant (see multiple comparisons/ Bonferroni correctio)

Your Likert scale is small. 4 , the question is average of how many questions (for the same subject) create the DV?
If you decide to do regression you should read about Likert scale linear regression or ordinal regression