I performed a linear regression, in which I got a negative slope coefficient. May I interpret as if the explanatory variable increases, then the response variable decreases?
As Roberto Molteni says, the interpretation is okay. But, before fitting a linear regression we should see if there is a trend for a linear relationship in the scatterplot. In your case, this relation is not present. Also, I am afraid your slope is not significantly different from zero, have you checked this?
As Roberto Molteni says, the interpretation is okay. But, before fitting a linear regression we should see if there is a trend for a linear relationship in the scatterplot. In your case, this relation is not present. Also, I am afraid your slope is not significantly different from zero, have you checked this?
Relation seems insignificant from the graph and the -0.06 effect is simply negligible or you can say insginficant, please look at its Beta's P-value, if its more than 0.05, it will be called as insgnificant.
According to the attached chart I don't see any dependency between variables in your particular model. Also the r2 (R square) value is very low (0.02). R2 = 0 means there is no dependency, R2 = 1 means absolute dependecy (simplified interpretation). You don't show us p-value, but if it is higher than 0.05, your model is statistically insignificant..
Yes, your interpretation is right in terms of the relationship between your explanatory and response variable. Since it is a linear regression, then you may interpret it this way: a unit increase in x decreases your y by 0.06. However, it seems that the coefficient estimates are not significant. When interpreting the regression coefficients, it is important that they are statistically significant from zero (i.e., p-value < the level of significance). Otherwise, we cannot interpret them. Also when using linear regression (especially when estimating it using OLS [ordinary least squares]), please be cautious of outliers as it may affect the estimation of your parameters (or slope). I suspect that you have 2 outliers, i.e., two data points on the left side of the plot.
Here are some model diagnostics you can check:
1. Normality of error;
2. Homoscedasticity;
3. Non-degrading collinearity (i.e. VIF < 10);
4. Non-biased estimates (due to suspected outlier);
If you satisfy these four and your coefficient estimates (or slopes) are significant, then you can proceed in interpreting.
Hi! Be aware because your model is not statistically significant. The R Square is too low and show you that the model do not explain the relationship between the data. The list that Paul Zython recommends to you is very useful to check the results of your model before you proceed to interpreting.