Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

# How to interpret bgf.summary

See original GitHub issue

1. How to read `bgf.summary`? what does each column and row mean?

coef se(coef) lower 95% bound upper 95% bound
0.271124 0.003597 0.264073 0.278175
156.815975 3.151111 150.639797 162.992154
0.331748 0.014789 0.302761 0.360735
1.176144 0.061768 1.055079 1.297209
2. I’ve tried to change the `penalizer_coef`, below is set at zero, the more I increase it, the worse it is.

I need some guidance on closing the model’s prediction gap. Which part do most beginners get wrong?

Thanks

### Issue Analytics

• State:
• Created 4 years ago

1reaction
psygocommented, Nov 18, 2019
1. How to read bgf.summary? what does each column and row mean?

The most important part is the first column, the coefficients themselves. The other columns refer to the errors around their estimation. Imagine each coefficient is a Gaussian distribution: `se(coef)` is the standard error, the sigma, it gives you an estimate as to how good your model conversion was. The lower and upper bounds are the range of values the coefficients are assuming under 95% confidence. If the range is too high, you might want to check for better convergence.

1. I’ve tried to change the penalizer_coef, below is set at zero, the more I increase it, the worse it is.

Usually, the `penalizer_coef` can vary somewhat comfortably from `0.001` to `0.5` — though it would be nicer to have it at max around `0.1`. How high did you go? Too high and the model migh basically become useless.

Also, I would say that you should focus on making your model fit well with the `0` and `1` groups of the holdout vs calibration purchases graph, as they represent 70-90% of the customers usually. There are other graphs that will indicate your model’s performance also, try checking them out in conjunction with this graph.

0reactions
GrowthJeffcommented, Aug 6, 2020

Also, I would say that you should focus on making your model fit well with the `0` and `1` groups of the holdout vs calibration purchases graph, as they represent 70-90% of the customers usually. There are other graphs that will indicate your model’s performance also, try checking them out in conjunction with this graph.

I found this very insightful. Thank you!

#### Top Results From Across the Web

The BFG Summary
As the book starts, a young girl named Sophie lies in bed in an orphanage. She can't sleep, and sees a strange sight...
The BFG Summary
Our story begins during the Witching Hour, a time of night when humans are asleep and creatures from the shadows get to roam...
We can visualize this relationship using the Frequency/Recency matrix, which computes the expected number of transactions an artificial customer is to make in ......
The BFG Summary
Thanks for exploring this SuperSummary Plot Summary of “The BFG” by Roald Dahl. ... The BFG for example, cannot read, but has skills...
The BFG Summary & Study Guide
The BFG Summary & Study Guide includes detailed chapter summaries and analysis, ... comprehensive information and analysis to help you understand the book....

#### Troubleshoot Live Code

Lightrun enables developers to add logs, metrics and snapshots to live code - no restarts or redeploys required.
Start Free