Hi Deana,

Could you please make your question more clear.

From what I have understood from your question, you are trying to do a logistic regression. As Sofia mentioned above, it is appropriate when you have a binary outcome (the response either takes one of the two values: For example 1 or 0, Yes or No...).

The regression model (linear regression, logistic regression) is trying to  use  availabe variables (covariates) to understand about the response variable (Y). We have two general goals:

1. Making inference: To see how the covariates related to the response, for example: If an employee has a higher education degree, how likely he/she has high salary (high salary is defined as having earned more than 8000Euro per month for example)...

So you really want to see how the respose changes with regard to those covariates that we think (or we have evidence to say) that they are related to the response.

2. Making prediction: We build a model based on available data. Based on these models (assumed that the assumptions under the model are satisfied), we want to make prediction on the response based on the "new" values of the covariates (pay attention that these new values should be in the range of those in your data so as to avoid extrapolation) as it is not alway easy to measure Y.

For example: You build a model to predict how likely a patient within 48 hours after a stroke survive, given the measurements of their blood pressure, PbO2, HbO2, SvO2...

So you collect data based on those subjects in your study, then you construct a model, check the reasonability of the model and use this model for the future patients to predict their survival ability.

Come back to your question, it is possible that there is only one covariate (X) that the scientist think that it might be strongly related to the response Y. Before going to fitting model step, some exploratory steps could be done to see how the data reacts.

Related to your research question, it is not clear for me, so I can not say anything about it.

All the best.

Hi Thao,

Thank you for your explaination... here are what i trying to do..

I try to use a logistic regression on my analysis.. so my data is that i try to look how my amount of rainfall(mm) in each years is reacted with number of raindays. So it is posible for me to use logistic regression.. as from your explaination above, i think my cases is no 1.

Both your variables are continuous, so you can't use logistic regression. For these data you should use linear regression.

Hi Deana,

Stephen is correct.

When you have your response variable as a continuous (measurements take on many possible values in the real range) then you could consider a simple linear regression (when you have one covariate) or multiple linear regression (when you have more than one covariate).

Logistic regression is used when your response is binary (takes one one of two values: 1/0; Yes/No; Agree/Not agree for example).

Thao.

Hi Thao,

im still wondering. ok lets say, how about if i want to know is Number of Rainy days will have a significant relation with Amount of Rainfall (mm) then it is possible or suitable to use Logistic Regression?

No; again, for continuous dependent variables you should use linear regression.

If number of rainy days is an ordinal categorical variable (i.e., if there are only a few levels, like {1, 2, 3}) and it's the dependent/outcome variable, then it may be appropriate to use ordinal regression. But as Thao explained above, you use linear regression for a continuous outcome variable.

Tq Stephen. now i understand. tq for your help :)