# Do we have any assumption on the distribution of Data for linear regression? [duplicate]

I am in general confused about the assumptions of linear regression and on whom are those assumptions taken to be and what are the consequences of these assumptions on the validation procedure for the regression model. Can someone help in this regard.

• We have quite a few posts that discuss assumptions for linear regression. It might help clarify what you need if you begin by reading some of those. Are you talking about ordinary multiple regression or linear models more generally? Do you mean to include any linear model of the form $E(Y|X) =X\beta$ (which would include, for example L1 regression as well as Poisson or Gamma regression with identity link) or do you mean explicitly to restrict yourself to least squares? Do you mean in terms of fitting or are you talking about inference (hypothesis tests, confidence intervals etc)? May 25, 2017 at 4:53
• Posts you might find helpful in finding a specific question to ask that's not already here: 1. Where do the assumptions in linear regression come from? ... 2. Regression when residuals are not normally distributed .... 3. residuals are normally distributed but Y is not May 25, 2017 at 5:30
• 4. transformation to normality of the dependent variable in regression ... 5. how does linear regression use the normal distribution ... 6. independence of error in linear regression .$\qquad$ searches like assumptions linear regression will find more May 25, 2017 at 5:30
• If you want something different from the various pieces of linked information, please edit to clarify what you need (or post a new question) May 25, 2017 at 5:45
• @Glen_b Great link to a well phrased, conical question. Some student confusion comes from searching for the set of assumptions but there isn't a single set of assumptions. OLS is more like a hammer, and there are different situations (beyond hitting a nail) where a hammer does something useful. May 25, 2017 at 6:39