Respected all,
I want to clarify,
1) can I use the chi-square test to check the association between gender and Retirement confidence ( measured as Yes, No, Not calculated)
and Gender and monthly savings ( measured as 0-20%, 20-40%, more than 40%).
and after getting a significant p-value, interpret the result with cross-tabulation?
2) Is this necessary to form a hypothesis for the chi-square test?
Because some previous literature (emerald, Elsevier) use only Research questions, while some other set Hypothesis before applying the Chi-square test.
3) Same confusion regarding the use of chi-square vs logistic regression?
I saw some papers where researchers use chi-square and some use regression (in the case of categorical variables).
please clarify this doubt
1) yes.
2) yes. A hypothesis test neccesarily tests a hypothesis. The hypothesis tested by the usual Chi² test applied to contingengcy tables is that the row- and column frequency distributions are independent. A reasearch question usually is that there is some dependency, and the hypothesis test is done to check if the observed data provides enough information to reject the hypothesis of independence. Such research questions are often not very helpful, as the interesting thing would be a particular structure of the dependency, which is usually not adressed.
3) a Chi² test is an approximate likelihood-ratio test for hypotheses about coefficients in a linear model: under the tested hypothesis, the log likelihood ratio is approximately Chi² distributed. If the response variable has a distribution with variance independent of the expectation, the approximation can be very bad for small samples, and the F-distribution is a better approximation. When the response has a normal distribution, then the log likelihood ratio has an exakt F-distribution under the tested hypothesis. For counts, the response distribution can be assumed Poisson, and the Chi² approximation is the best we have. Same for a binomial response. The statistical model behind these tests (Chi²- or F-tests, that can be formulated as z- and t-test when there is only a single coefficint under consideration), can be described as linear models (aka regression models), in which coefficients are tested. For a Poisson variable such models are called log-linear models, for binomial variable they are called logistic models, and for a normal variable such models are just called (linear/regression) models. In simple cases the model is obvious and does not need to be stated explicitely and one just refers to the (only meaningful) test in this model, what can be a t-test, a z-test, an F-test or a Chi²-test.
Thankyou sir for reply @Jochen Wilhelm
1) ok I go with chi-square test
2)
In my study, i use chi square test to find association between
4 factors (gender, age, income, and employees covered under different pension schemes) all measured on the categorical variable
and
(1) retirement confidence ( Yes, No, Not calculated), (2) monthly savings behaviour (0-20%, 20-40% and more than 40%), (3) monthly additional contribution in the provident fund (less than 5%, 5-10%, 10-15%, and more than 15%) and (4) withdrawal from Provident fund (Yes, No).
then I need to make 16 hypotheses...
Can we set such a large number of hypotheses for a research paper?
and it is the right way to set a hypothesis?
H01: There is no significant association between gender and retirement confidence
H02: There is no significant association between age and retirement confidence
.....and so on
Hello Aarti,
Yes, you may evaluate as many hypothesis tests as are germaine to your research. However, do recognize that, the more tests you run, the greater the aggregate risk of at least one type I error (false rejection of a null hypothesis). To compensate for this, you could evaluate each hypothesis at a more stringent (lower value for) alpha level. There are well-known methods for this, here are two:
1. Bonferroni adjustment (simple to do; suffers in power relative to some other methods)
2. Benjamini-Hochberg method (involves sorting results based on computed p-values, then using a sliding scale for significance based on the rank).
The R function, p.adjust, offers these and a host of other adjustments as well: https://www.rdocumentation.org/packages/stats/versions/3.6.2/topics/p.adjust
Yes, your hypothesis examples are fine.
Good luck with your work.
@David Morse
Thankyou sir
I do the same
And also check these two methods
And then if there is some problem, then I ll ask you.😊
Thankyou
A chi-square test is a statistical test used to compare observed results with expected results. The purpose of this test is to determine if a difference between observed data and expected data is due to chance, or if it is due to a relationship between the variables you are studying. Therefore, a chi-square test is an excellent choice to help us better understand and interpret the relationship between our two categorical variables.
To perform a chi-square exploring the statistical significance of the relationship between s2q10 and s1truan, select Analyze, Descriptive Statistics, and then Crosstabs.
Find s2q10 in the variable list on the left, and move it to the Row(s) box. Find s1truan in the variable list on the left, and move it to the Column(s) box. Click Statistics, and select Chi-square.
Click Continue and then OK to run the analysis. Your output should look like the table on the right.
- Badges
- Science topic
- Similar topics
- Counseling Psychology
- Confusion