@Hooman Logistic regression would be appropriate as your dependent variable is still binary not continuous. For example; you had responses for Item 1= 2; Item 2 = 1; Item 3 = 2; Item 4 = 2 and thus, mean response would be 7/4 = 1.75 which is still in the range of 1-2 (No/Yes). Logistic regression would take cut point at 1.5 (as you coded 1 for No and 2 for Yes) and software would take value below 1.5 as No and above 1.5 as Yes. Had you coded 0 for No and 1 for Yes in your dependent variable, then cut point would have been 0.5.

You do not have a continuous scale. It is discrete with values 4, 5, 6, 7, or 8.

You haven't said if the responses these yes/no questions are supposed to form a single dimension or if you explored this (e.g., with IRT). You haven't said if you are interested in the bivariate association or conditional one. Or interactions.

I recommend deleting this question and writing one with enough information for people to give informed suggestions.

No, you can't run a linear regression since your independent variables are discrete but if your objective is to test the linear association I.e to see whether change in the level of the independent variables will bring about a change in the level of the dependent variable then you need to transform the independent variables into continuous by computation if the order of categorization is scored but if otherwise, you might have just left the outcome variable in dichotomous form (Yes/No) and go for ordinal logistic regression considering your ordinal independent variables.

Ismail Bamidele Afolabi Thank you for your response. Yes I did that and I transform the items of my independent variable into a continuous value. This should be fine right?

step wise regression is a method of fitting regression models in which the choice of predictive variables is carried out by an automatic procedure. In each step, a variable is considered for addition to or subtraction from the set of explanatory variables based on some prespecified criterion

You might be okay treating your dependent variable as if it were a continuous variable (and conditionally normal ! ). But, as has been suggested, you are probably better off with logistic regression or ordinal regression. Some implementations of logistic regression allow you to easily define your dependent variable as positive counts out of some possible counts. (Like in R, you can use a matrix of successes and failures as the dependent variable). But consider the question asked by Daniel Wright : Does it make sense to combine these four questions into a single scale ?

This model type necessitates the use of a binary dependent variable. A binary variable may take just two values, for example, pass or fail. For instance, political scientists use stock market performance to estimate the chances of the current President of the United States gaining reelection.

Logistics regression or decision trees would be two good options.

Your p-value reflects the credit of the null. If you test for B=0 then the transformed outcome associated with your independent variables not with original outcome variables.

Try Logistic regression