My independent variable is categorical with 3 categories and dependent variable which is continuous (GPA). Then I want to test if these two are related. Is the chi-square test for independence valid for this? If it is valid, how do I recode variables in stata?
The Chi²-test test the association between two categorical variables. Here you have
The usual test that is applied here is the F-test (ANOVA).
If you insist on the Chi²-test then you can categorize the values of your dependent variable. However, I would not recommend this because you irgnore a lot of information that comes with the continuous data.
Hi Jane,
The Chi-squared test tests whether there is a different between observed and expected values.
If you are interested whether the means across all three categories are statistically different then you would benefit from using an ANOVA. ANOVA is also a suitable extension of Chi-square with more than two groups.
E.g., does GPA vary depending on socioeconomic background?
http://en.wikipedia.org/wiki/Analysis_of_variance
Also, Google "More than two groups: ANOVA and Chi-square", there is a Stanford PowerPoint which will help you.
Best regards,
Dan
Not valid, chi square is only for categorical variables, you should recode your variable GPA, creates a new variable: generate new=0, then replace new= 1 if GPA> 1 & GPA 11 & GPA 21 & GPA
Hi Jane,
Here's how you do it in Stata. Lots of helpful advice at
http://www.ats.ucla.edu/stat/stata/whatstat/whatstat.htm
* -------- ----------------------
* GPA is continuous (the outcome)
* x1 is the independent variable (categorical, 3 levels)
* Use the auto data set as an example
sysuse auto , clear
gen x1 = rep78 if inlist(rep78,1,2,3)
gen GPA = price
keep GPA x1
drop if x1==.
* different ways to do the analysis
anova GPA i.x1
reg GPA i.x1
kwallis GPA , by(x1) // non-parametric
* -------- ----------------------
Good luck
Rich
Dear Jane
Jochen is right.
If GPA has strongly non-normal distribution, You can use non-parametric ANOVA (Kruskal-Wallis).
If categorial predictor is ordered - tell about it. There are some another test in that case.
I would keep GPA continuous. If you use one of the methods suggested above, you can add in more factors to your analysis. Do you think age, gender, coursework, major, etc, have an effect? A Chi2 would have a difficult time with this type of analysis. The methods suggested above a good for that.
you could for Spearman Rank correlation whenone is qualitative and the other one is quantitative
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis