Tweeted twitter.com/StackStats/status/885304962711814144
2 deleted 159 characters in body
source | link

There are several posts about how to select features. One of the method describes feature importance based on t-statistics. In R varImp(model) applied on linear model with standardized features the absolute value of the t-statistic for each model parameter is used. So, basically we choose a feature based on its p-value. However, p-value (together with t-statics) is based on the null hypothesis that a coefficient deviates from zero andstatistics, meaning how precise it is the coefficient. ItBut does not tell me how my model performs without the feature, so it does not tellpreciseness of my coefficient tells me anythingsomething about the predictive abilities of the feature. ?

So,Can it can happen that my feature has a non-significant plow t-valuestatisstics but would still improve (lets say) accuracy of the model.

Question: When? If yes, when would one want to exclude variables based on the p-values (or t-statistics)? Or does it give just a start point to check the predictive abilities of non-significantimportant variables?

There are several posts about how to select features. One of the method describes feature importance based on t-statistics. In R varImp(model) applied on linear model with standardized features the absolute value of the t-statistic for each model parameter is used. So, basically we choose a feature based on its p-value. However, p-value (together with t-statics) is based on the null hypothesis that a coefficient deviates from zero and how precise it is. It does not tell me how my model performs without the feature, so it does not tell me anything about the predictive abilities of the feature.

So, it can happen that my feature has a non-significant p-value but would still improve (lets say) accuracy of the model.

Question: When would one want to exclude variables based on the p-values (or t-statistics)? Or does it give just a start point to check the predictive abilities of non-significant variables?

There are several posts about how to select features. One of the method describes feature importance based on t-statistics. In R varImp(model) applied on linear model with standardized features the absolute value of the t-statistic for each model parameter is used. So, basically we choose a feature based on its t-statistics, meaning how precise is the coefficient. But does the preciseness of my coefficient tells me something about the predictive abilities of the feature?

Can it happen that my feature has a low t-statisstics but would still improve (lets say) accuracy of the model? If yes, when would one want to exclude variables based on the t-statistics? Or does it give just a start point to check the predictive abilities of non-important variables?

1
source | link

Is it wrong to choose features based on p-value?

There are several posts about how to select features. One of the method describes feature importance based on t-statistics. In R varImp(model) applied on linear model with standardized features the absolute value of the t-statistic for each model parameter is used. So, basically we choose a feature based on its p-value. However, p-value (together with t-statics) is based on the null hypothesis that a coefficient deviates from zero and how precise it is. It does not tell me how my model performs without the feature, so it does not tell me anything about the predictive abilities of the feature.

So, it can happen that my feature has a non-significant p-value but would still improve (lets say) accuracy of the model.

Question: When would one want to exclude variables based on the p-values (or t-statistics)? Or does it give just a start point to check the predictive abilities of non-significant variables?