If you please see more details about logistic regression in my website. scroll down the page then you can see the topic in English language.

http://www.alghamdi-biostatistics.com/logisticregression.htm

Hello Ashwini. Basically, multivariate statistic is any kind of analysis that use more than 2 predictors and more than 2 criteria, in one analysis. It means that you have many different elements that help you to predict others. Examples of those are multivariate analysis of variance (MANOVA), and structural equation modeling (SEM). The logistic regression is considered like one of them, but, you have to use one dichotomous or polytomous variable as criteria. The basic assumptions for multivariate analysis are linearity, homocedasticity, and correlation but not multicolinearity between the varaibles... and others that are dependent of the kind of analysis.

Logistic regression measures the relationship between a categorical dependent variable and usually a continuous independent variable (or several), by converting the dependent variable to probability scores. It is used extensively in medical and social sciences fields.

Thank you for the helpful discussion Dr. Miljail and Dr. Mohammed. To be precise, i was interested to know more about the conditions in which a multivariate analysis are suitable, and how do we set up a regression equation and what are the requirements for the multivariate analysis. Only in the follow up (interventional study) or having more than two independant variables or dichotomous variable? Also, I was interested to know about setting a regression equation for multivariate and logistic regression analysis. Let us consider an example of micronutrient deficiency in a population. What are the tests need to be done after identifying the dependent variables and the independant variables (eg. socio-demographic variables) in a categorical data? Suppose I have a done a Chi-Square test to establish association, and logistic regression test to know the estimated risk. Then for what kind of study data needs mutivariate analysis. What are the basics of of this test to do in a SPSS, or do we need other special statistical tools? I would be thankful if you suggest further links for detail information. Thank you

Depends on whether one wants to carry out principal component analysis, or cluster analysis. In the former we want to reduce dimensionality in the data i.e. which variables are important in causing variation in the data. While in cluster analysis we want to group our data based on some variables.

Multivariate analysis is about finding an effect size of a singe independant variable in the presence of actual or potential confounding. Often, it will be used for multiple independant variables, the so called "shotgun approach", by dumping large numbers of variables as independant variables into an equation, and then interpreting the effect size for each of these independant variables.

Generally, your study should have a single independant variable of interest, a series of potential confounders, and the dependant variable. As long as you don't have any imbalance in the independant variable across a range of potential confounders, technically, you do not need a multivariate analysis. This is why high quality RCT's do not use multivariate analysis, because the randomization process ensures balance across a range of confounders.

I'm not shure if you should try to run this kind of analysis by your own... There are several pitfalls in this models and you should ask for professional help to run these models...

I know that it could sound prepotent from me, but if you put your data in a software and press "run", it will give you a result, but it could be completely wrong...

Thank you Dr. Nicolas, Dr. Mohammad and Dr. Sandro for information. Dr. Sandro, I will definitely contact to statistician if I need to run these models on my data. Thank you once again.

I would like to stress the role of studying interaction between different covariates in multivariate analysis. It could be useful for you to delve the role of effect modifiers.In most applications, statistically significant interactions between covariates and a main predictor are qualifications or extensions of hypothesized associations that merit

careful representation in a biomedical research study.

If you please see more details about logistic regression in my website. scroll down the page then you can see the topic in English language.

http://www.alghamdi-biostatistics.com/logisticregression.htm

Can we consider ordered or multinomial  logistic regression as multivariate ?