Hi,
I am new to a quantitative research and wanted to use a logistic regression to count the probability. However, I found that my data is not normally distributed and therefore should use a non-parametric test. I used SPSS but willing to learn any new software if I could count my non parametric data to obtain the probability result. Is there any software that could help solve my problem or I should just ignore the fact that my data is non-parametric?
You need to analyze your anova and counts on f significance .Wish you luck.
Logistic regression does not assume normality. You can conduct a logistic regression with these data in SPSS.
To be clear, logistic regression does not simply calculate probabilities but compares relative probabilities for some (discrete) outcome. I suggest reading up on odds ratios before conducting the test.
Medhat Elsahookie ANOVA will not work with count data (i.e., proportions).
Non-parametric data can be measured using logistic regression. Logistic regression is a parametric test, meaning it is a set of distributions that a finite number of parameters can parameterize. To measure non-parametric data with logistic regression, the optimal bandwidth value must first be determined based on the maximum Cross Validation value. Alternatively, when covariates are measured with error, GMM nonparametric correction methods can also be used for logistic regression. When nonlinear relationships exist and transformation does not linearize them, nonparametric methods should be used if the sample size is large enough.
I suppose I have to add the obligatory comment that there is no such thing as non-parametric data.
There are parametric analyses and non-parametric analyses *, that handle data in different ways. But the data itself can't be described as parametric or non-parametric.
Anggara Raharyo , while you received some useful advice in this thread, it would be helpful if you describe the nature of your dependent variable. (Is it in fact binary in nature ?)
In general, whether the data are normally distributed or not has nothing to do with what analysis is appropriate.
__________________
* Also with the caveat that there's not always a clear demarcation between parametric and non-parametric analyses. But there's a useful tradition there.
Non-parametric tests cannot estimate probability like logistic regression, but they can examine the significance of differences or relationships between variables.
Non-parametric tests produce a test statistic and p-value, which can be used to establish significance.
If your data is not normally distributed, employ a non-parametric test. Always evaluate the assumptions of the statistical procedures.
R and SAS have big user communities and many resources for learning new software.
Chi-square test is a none-parametric test, you can use it for a categorical response meaning binary or multinomial dependent variable.
Why should your data be normally distributed??? For regression it is better if they are evenly distributed. You can try your logistic regression easily with this software:
www.lerenisplezant.be/fitting.htm
Let me know if you need assistance.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Artificial Intelligence