Which statistical test is used to analyse cause and effect relationship between two variables independent of any grouping variables? To get a more clear picture of this, let me illustrate it with an example. There are two variables A and B. I want to test if A has an effect on B or B has an effect on A. For instance, we know that engagement and learning achievement effect each other. Like if students are highly engaged, this will lead to better learning achievement of such students. I want to check similar sort of relationship.
There is no statistical analysis, by itself, that will demonstrate a cause and effect relationship. You must also consider the research design. Can you perform an experiment in which subjects are randomly assigned to Variable A and perform analysis to test whether Variable A assignment is related to outcome Variable B at follow-up? This randomized controlled trial (RCT) design is the most robust for establishing cause-and-effect. If not, the next robust design is a longitudinal design, in which you collect data on Variable A and B repeatedly over time for each subject and perform an analysis to see whether subjects that experienced an increase in A also showed an increase in B (either in real time or as a lag function), or some other hypothesized relationship. A bidirectional effect of the kind you describe is even more difficult to demonstrate. It cannot be demonstrated simply by examining correlation, regression or other statistical analysis.
Amna:
I recently asked a similar question and got some very detailed answers, so I am sharing the thread with you:
https://www.researchgate.net/post/Are_the_use_of_regression_and_correlation_coefficients_enough_to_measure_effect_of_the_independent_variable_on_the_dependent_variable
Best regards,
Debra
I think you must consider t-test of independent samples if-and-only-if your data set is on the continuous scale, i.e. at least not falling short of the interval measuring scale, and observing the following assumptions: if independent observations are obtained from a normally distributed population, and also knowing that the two samples from which data was observed are homogeneous in variance. Else, you will have to lookup other nonparametric techniques if these as well as other parametric assumptions fail. Let me know if this helps or there are other issues to consider in your study.
Get your hands on :
Kothari, C.-R., (2004). Research Methodology. Methods and Techniques (second revised edition). New Age International Publishers.
The answer to your question is in there somewhere... I believe.
It seems to me that both your variables are dependent... what leads on student engagement? It seems to me that student engagement and learning achievement must be correlated, but they depend on something else, don’t they?
There is no statistical analysis, by itself, that will demonstrate a cause and effect relationship. You must also consider the research design. Can you perform an experiment in which subjects are randomly assigned to Variable A and perform analysis to test whether Variable A assignment is related to outcome Variable B at follow-up? This randomized controlled trial (RCT) design is the most robust for establishing cause-and-effect. If not, the next robust design is a longitudinal design, in which you collect data on Variable A and B repeatedly over time for each subject and perform an analysis to see whether subjects that experienced an increase in A also showed an increase in B (either in real time or as a lag function), or some other hypothesized relationship. A bidirectional effect of the kind you describe is even more difficult to demonstrate. It cannot be demonstrated simply by examining correlation, regression or other statistical analysis.
Personally I cannot understand your question because:
1. cause and effect is talking about independent and dependent variables.
2. two variables independent of any grouping variables cannot be related to cause and effect
3. two independent variables can be cause and cause
4. two independent variables cannot be effect and effect since they are independent,
5. best tool can only be determined when the type of relationship is known or established.
6. however, check Pearson product moment correlation for it can be used to correlate A:A, A:B, B:B and B:A.
I would recommend that you use SEM model although you can also use the SEM PLS model that is useful in predictive models as the example you mentioned of evaluating the effect of 2 independent variables correlated to a dependent variable.
I would go for a regression analysis or an SEM. I personally would use t tests to establish that a difference exist in the outcome variable if the independent variable has two levels ( like treatment and control, exposed non exposed). Then further evaluate the relationship by a regression analysis. However consideration of study design is important. A prospective or a retrospective cohort would give a better indication of a cause and effect relationship than a case control. A Randomized control trial is robust in establishing cause and effect relationship.
I think checking conditional probabilities is a good way to check causal effect. If your population doesnt come from normal distribution , you can give a prior to get more suitable results with bayesian approach.
Correlation analysis is mostly conducted to determine if a relationship exists between variables. Regression analysis is used to determine the effect of one variable on the other. The technique to be used is based on what the researcher is looking at.
How can it be possible sir?
I have three IVs
IV: Training design, Trainee motivation, Organizational support
DV: Training transfer
I want collect IVs data just immediate after training and DV data after 3 months of training. What procedure can i take sir?
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis