The logistic regression (LR) model is a commonly used mathematical method to establish the relationship between conditioning factors and landslides . Some studies have compared the logistic regression model with other methods and have found logistic regression to be the most accurate of these techniques. Therefore, a forward stepwise logistic-regression model is applied to create landslide-susceptibility maps. The predicted values range from 0 to 1, and can be defined by the following formulas: y = b0 + b1x1 + b2x2 +...+bnxn and p = ey/1 + ey

where y is the linear logistic model, b0 is the intercept of the model, n is the number of landslide-controlling factors, b is the weight of each factor, x represents the landslide conditioning factors and P is the probability of landslide occurrence

Two important parameters, tolerance (TOL) and the variance inflation factor (VIF), were used to determine the colinearities between independent variables. If TOL < 0.4 and VIF > 2, those parameters were not selected in the logistic regression models. In this study, there are no high correlations among the dependent variables. The training datasets were then used in the SPSS software to model the relationship between the probability of landslide occurrence and the conditioning factors. The last category was assigned as the reference category for each categorical variable. The spatial relationship between landslide occurrence and conditioning factors was assessed using the LR model with different training datasets.

Sujan's answer is excellent, and comprehensive. I'd note as well, since you want something basic and practical, that linear and multiple logistic regression can be done through a number of analytic programs. Excel, for example, has tools for this. So do most statistical packages. ArcGIS has, I think, a weighted regression program in its statistical tool kit. And if you're using or have available SPSS or SASS, well, the same. So I think with an undertsanding, provided by Sujan, you should look at the program available, query multiple regression, and then see if that serves.

@Sujan, well the explanation provided by you is good and a new thing in this explanation is Tolerance and Variance inflation factor which is not explained in the research articles. But my question is that how to analyze it because I have the knowledge about software and formula but how they did it. for example in frequency ratio method we compare the landslide pixel with each class of each parameter and get the frequency values. The same how we do in logistic regression model because some researchers said that convert the raster file to ASCII format and use in SPSS or any statistical software but what are the fields to use for analysis and researchers also used non landslide area while analyzing landslide with logistic regression model but not any other techniques/model that is confusing, so can you please clarify it to me.

Hi @Geraud Poueme Djueyep

The file is very good and informative until

I have only the inventory of landslide not any sub category of it, is it necessary to have the non landslide area to be marked randomly in the study area?

Bi = presence of a potential landslide conditioning factor,

Bi = absence of a potential landslide conditioning factor,

S = presence of a landslide, and

S = absence of a landslide

The presence means that the conditioning factors which we used in our study i.e. slope aspect etc, and presence of landsdlies means landsdlie inventory but what does the meaning of absence I did not get it.

Which one we have to use from them?

W+ =ln((npix1/(npix1+npix2))/(npix3/npix3+npix4))

=lng((npix1*(npix3+npix4)/((npix1+npix2)*npix3)

W- =ln((npix2/(npix1+npix2))/(npix4/npix3+npix4))

=ln((npix2*(npix3+npix4)/((npix1+npix2)*npix4)

these are really confusing for me..

Wmap = Wplus + Wmintotal – Wmin

C =W - W + -