OLS (Ordinary Least Square) Assumptions

• A Linear Regression is Linear because the Equation is Linear.
• To verify the Linearity between Independent and Dependent Variable apply Scatter plot.
• If the Data points form a pattern that looks like a Straight Line then Linear Regression is Suitable.

If the pattern doesn't looks like a Straight Line, then we need to apply

• A Non Linear Regression or
• Exponential Transformation or
• Log Transformation.

• Technically, Endogeneity occurs when a predictor variable (x) in a regression model is correlated with the error term (e) in the model.
• This usually occurs due to the Omitted Variable Bias (when we forget to include a relevant variable in the Model).
• As the thing that we don’t explain with our model goes into the error, the error would be correlated to the variable that we have omitted.
• Omitted Variable Bias is hard to find. When in doubt, include the Variable in the model and try it yourself. 

• We need to consider that our Error Term is normally distributed.
• We only need Normal Distribution while making Statistical Inferences.
• T-tests and F-tests work because we have assumed normality.
• If Error Term isn't normally distributed then we can use Central Limit Theorem.

• If the Mean of the Error Terms is not expected to be zero then the line is not the Best fitting one.
• Having an Intercept Solves the problem. 

• Homoscedasticity means to have “Equal Variance”. The error terms should have equal variance with one another.
• When the error terms don’t have “Equal Variance”, then Heteroscedasticity happens.

In order to prevent Heteroscedasticity, we need to

• Look for Omitted Variable Bias.
• Look for Outliers.
• Apply log Transformation.

• Autocorrelation is a mathematical representation of the degree of similarity between a given time series and a lagged version of itself over successive time intervals.
• It is similar to the Correlation between two different Time Series but Autocorrelation uses same time series twice.
• One is Original form and other one is Lagged version of Original form.

To detect autocorrelation

• Plot all points and check for patterns or
• Use Durbin - Watson test.

There is no remedy for Autocorrelation. Instead of linear regression, we can use

• Autoregressive Models.
• Moving Average Models.

• We observe Multicollinearity when two or more Independent Variables in a model are “Highly Correlated”.
• Correlation could be Positive (or) Negative.
• This will lead result in less reliable predictions.
• Multicollinearity can be noticed easily by finding the correlation between each two pairs of Independent Variables.

To fix Multicollinearity: 

• If the Dataset is small, we can drop one independent variable. Or we can Transform them into One Variable (Eg. Average of two Variables)
• If the Dataset is large, we will use Ridge and Lasso Regression.