91Ó°ÊÓ

In simple linear regression, there are a few essential assumptions that we need to check to ensure the model is appropriate for our data. These assumptions include:

• Linearity: This assumption implies that there’s a linear relationship between the independent variable, X, and the dependent variable, Y. Simply put, changes in X lead to proportionate changes in Y.
• Independence: The observations should be independently randomly sampled. No two observations should be related or influence each other.
• Homoscedasticity: This means that the residuals (errors) should have constant variance at every level of X. In other words, the spread of the residuals should be the same regardless of the value of X.
• Normality: The residuals should be normally distributed, which can often be checked using a QQ-plot.

Checking these assumptions helps ensure that the predictions made by the model are reliable and valid for interpretation.

The coefficients in a simple linear regression model include the slope and the y-intercept. They are significant because they describe the relationship between X and Y. - **Slope ( b ):** This value represents the change in Y for each unit increase in X. A positive slope means there is a positive correlation between X and Y, while a negative slope indicates an inverse relationship. - **Y-intercept ( a ):** This is the estimated value of Y when X is zero. Although sometimes X cannot realistically be zero, the y-intercept provides a baseline level of Y from where changes begin. These coefficients enable us to create the regression equation: Y = a + bX . This equation helps in predicting the value of Y for any given X within the range studied. Calculating these coefficients involves using mathematical formulas, where sums of products and squares of the variables play a crucial role.

A scatterplot of residuals is an essential diagnostic tool in evaluating a regression model. Residuals are the differences between the observed values of the dependent variable and the values predicted by the model. When creating a scatterplot of residuals versus fitted values (predicted values), we are looking for:

• No apparent patterns: This suggests that the model's assumption of linearity and independence of errors holds.
• Evenly spread residuals around the horizontal zero line: This implies homoscedasticity, confirming that the errors have constant variance across levels of the independent variable.

If the scatterplot shows any distinct pattern, such as a funnel shape or clear curvature, it may indicate that the simple linear regression model is not appropriate, and further investigation or transformation of variables might be necessary.

Two critical assumptions of regression models are homoscedasticity and normality of residuals.

Homoscedasticity

This assumption requires that the residuals' variance remains constant at different levels of the independent variable. We can think of it as the 'spread' of errors being similar everywhere. In practice, this can be tested using a plot of residuals against fitted values. If the spread of residuals changes (e.g., fans out or funnels in), it violates homoscedasticity.

Normality

Normality means that the residuals (errors) of our model should follow a normal distribution. This can be assessed visually with a QQ-plot, where residuals are plotted against a theoretically normal distribution. If they plot approximately along a straight line, normality is in good shape. However, if points deviate significantly, it may suggest problems with this assumption. Ensuring these conditions are met is vital, as they affect the reliability of confidence intervals and hypothesis tests conducted on regression coefficients.