May I use Logistic regression Model with 200 sample size (100 organic farmers + 100 Conventional farmers). Dependent variable is "Adoption of organic farming (Yes/No) or (Organic/Conventional)" with seven Independent variables in five point scale ranging from Strongly Agree to Strongly Disagree. The objective of this study is to examine the influence of 7 different economic (Independent) variables in adoption of organic farming.
Thank you Dr. Senthilvel and Fabrice for ur response. In the given link
Sample size considerations:
Sample size calculation for logistic regression is a complex problem, but based on the work of Peduzzi et al. (1996) the following guideline for a minimum number of cases to include in your study can be suggested.
Let p be the smallest of the proportions of negative or positive cases in the population and k the number of covariates (the number of independent variables), then the minimum number of cases to include is:
N = 10 k / p
For example: you have 3 covariates to include in the model and the proportion of positive cases in the population is 0.20 (20%). The minimum number of cases required is
N = 10 x 3 / 0.20 = 150
If the resulting number is less than 100 you should increase it to 100 as suggested by Long (1997).
Hi,
Your question is very interesting one and important for all researchers.
What should be the minimum sample size for running logistic regression?
In your research you are having 200 samples. This is enough to do the analysis. No missing value in that. Ok.
Proceed your analysis. All the best for you.
Thanks.
.
rule of thumb with references in the link below :
https://www.medcalc.org/manual/logistic_regression.php
.
Thank you Dr. Senthilvel and Fabrice for ur response. In the given link
Sample size considerations:
Sample size calculation for logistic regression is a complex problem, but based on the work of Peduzzi et al. (1996) the following guideline for a minimum number of cases to include in your study can be suggested.
Let p be the smallest of the proportions of negative or positive cases in the population and k the number of covariates (the number of independent variables), then the minimum number of cases to include is:
N = 10 k / p
For example: you have 3 covariates to include in the model and the proportion of positive cases in the population is 0.20 (20%). The minimum number of cases required is
N = 10 x 3 / 0.20 = 150
If the resulting number is less than 100 you should increase it to 100 as suggested by Long (1997).
Dear Azam, for your study case, you have 7 independent variables with 5 point scales for your regression. In general, the minimum sample size is 3*7*5=105. Thus, I think 200 is enough. In addition, you divided evenly them into two parts: 100 organic farmers + 100 conventional farmers. It is fine for modeling.
Kelinger and Pedhazur recommend using at least 30 observations per independent variable with the condition of no less than 200 observations and no more than six hundred observations for the stability of Beta weights in regression analysis
I agree with Dr. Senthilvel
In your research you are having 200 samples. This is enough to do the analysis.
Just some additional points - you have obviously sampled on the outcome - otherwise you would not have got 100 in each group - presumably the organics farmers are over represented- this is what epidemiologists call a case - comparison (or a case control)l study - which is a good idea when the cases - the organics farmers are relatively rare . It means that you cannot estimate or discuss prevalence as you have controlled this by the design - that is how many organics farmers there are - without additional information
It would be useful to look at stuff on modeling case control data
Breslow and Day is the classic text
https://www.iarc.fr/en/publications/pdfs-online/stat/sp32/SP32_vol1-3.pdf
You need to be particularly careful if you have matched variables - eg a 45 year old male farmer in the cases and the controls- if this is the case - you cannot analyze age. because you have design that effect out To be clear - . matching can be a good thing if you already know that age is important and you want extra power to detect other effects.
https://www.researchgate.net/post/When_can_a_researcher_relax_the_rule_of_ten_events_per_variable_in_Logistic_Regression
This is what I have written on this design but there are many more detailed treatments
https://www.researchgate.net/publication/236671039_Epidemiology_An_Introduction
https://www.researchgate.net/publication/269846803_Health_Disease_and_Society_An_Introduction_to_Medical_Geography
Finally if you want to treat the predictors as ordinal then consider orthogonal polynomials instead of dummy variables for nominal variables - if you do not you will be estimating 4 parameters (and a base catheory) not 1 for each predictor and some of the other commentators have not appreciated this - see
https://www.researchgate.net/publication/234015221_Manual_supplement_for_MLwiN_Version_2.14
which has a worked example of fitting an underlying trend.
Edited to add - have a look at this where I discuss the 10 observations per predictor rule - I think it is too low in may circumstances
https://www.researchgate.net/post/When_can_a_researcher_relax_the_rule_of_ten_events_per_variable_in_Logistic_Regression
Data Manual supplement for MLwiN Version 2.14
Book Epidemiology: An Introduction
Article Health, Disease and Society: An Introduction to Medical Geography
- Badges
- Science topic