PakMediNet Discussion Forum : Biostatistics : SOME BIOSTATE TERMS
SOME BIOSTATE TERMS
Please explain the following:
1---What is the diference b/w Odd ratio and Adjusted-Odd ratio?
2---Difference b/w simple regression or step wise?
3---Is it necessary to conduct a univariate analysis first before logistic regression?
4---How variables are chosen to be entered in the Logistic Regression model?
5---what is Goodness of fit??
6---How to determine that the the logisric regression model fit the data?
Posted by: tariqaslPosts: 42 :: 20-03-2011 :: | Reply to this Message
Re: SOME BIOSTATE TERMS
1---What is the diference b/w Odd ratio and Adjusted-Odd ratio?
>>>Adjusted odds ratio is adjusted for confounders that were included in the model whic provided that odds ratio.
2---Difference b/w simple regression or step wise?
>>>simple regression is where there is one dependent variable and one exposure variable of your interest. For step-wise regression, Google it, cant write it in one sentence.
3---Is it necessary to conduct a univariate analysis first before logistic regression?
>>>No, but it is a good practice to do so
4---How variables are chosen to be entered in the Logistic Regression model?
>>> Based on biological basis and statistical significance. Biological basis is preferred followed by statistical significance.
5---what is Goodness of fit??
>>> A test to examine how well the statistical model fits your actual observed data
6---How to determine that the the logisric regression model fit the data?
>>> use Hosmer Lemeshaw test (there are others too, but none is as good as some tests for linear regression)
Posted by: rqayyumPosts: 199 :: 20-03-2011 :: | Reply to this Message
Re: Re: SOME BIOSTATE TERMS
Regarding adjusted odds ratio: it is automatically done when you are analysing multiple logistic regression model to adjust the confounding effect of the other variables.
Simple and stepwise regression: in simple regression model we have only on independent variable, and when we have more than one independent variables, we can use direct multiple regression analysis or by the use of stepwise regression model.for detail you may consult with some statistician or call me at (0300-4668681)
In logistic regression model we need at least dependent variable in binary mode.
regrading goodness of fit, it helps us to determine the agreement/association between expected data and observed data using chi-square test
how to determine that logistic regression fit the data: this question needs correction (not a valid question)
Uni-variate analysis is always done before any high level data analysis
Normal distributed data: we have certain methods to check the normality of data
first logical method: sample taken from BP patients , this is example of normal data, but sample taken from patients of rare disease is not normal
Posted by: ibrahim_apPosts: 138 :: 23-03-2011 :: | Reply to this Message