Please help me to understand the p-value concept in prospect of Multiple Regression (Backward Elimination). My understanding here is that
- The null hypothesis is we will assume that the dependent variable depends on all the independent variables.
- The alternate hypothesis is there is a significant chance that few or all of the independent variables has no significant impact on the value of the dependent variable
Now a significance value (alpha) means how much error (5%) we can tolerate and still say the null hypothesis is true whereas the p-value shares actual amount of error that our sample data is tolerating.
Suppose the case is P-value > alpha, so it means that our model is tolerating more than 5% error
Hence it clearly states that we have much more deviations from the null hypothesis and hence the alternate hypothesis is correct(i.e. if p-value is more we we rejecting the null hypothesis)
However from other sources i read that a null hypothesis should be rejected if the p-value is low.
Hence i am totally confused as what is happening. Please help.