Speed Up Stata

[Olivia Emma]

Hello

How we can speed up the Stata calculations please

[Jared Greathouse]

That, my friend, is indeed the question.

[Olivia Emma]

What is the answer please

[Daniel Schaefer]

What calculations? Post some code. Describe what you are trying to do.

[Jared Greathouse]

What is the answer please

Who knows? I sure don't. You post no code. You post no data. You ask a super general question, I give a general answer. I'm not even saying this to be mean. let's say i walked up to you on the street tomorrow. Let's say I asked you "How can we do the loop operation? Please help!!!!!" And then turned the microphone over to you. I provide no code. I provide no data. I just ask a question.

How would you answer me? How would you be ABLE to answer me? You couldn't. Again, I'm not saying this to be mean, but when you ask a question, always act like the question is being asked to you. I don't like Stack Overflow. In fact, I hate it a lot. But that is one good piece of advice they give, on question asking. Ask the question in the way you would want to be asked it.

[Nick Cox]

There are I think three answers to the question so far stated, but unfortunately I can't go beyond much the reactions to date.

1. Use a faster computer.

2. If you are not using the fastest implementation of Stata, get hold of it.

Manifestly, neither of those solutions is immediate or costless.

3. Tell us about your data and your code. It's possible in principle that you are doing something that could be speeded up with smarter code or a different data layout, but we can't tell without any details.

These may seem like weak, unhelpful replies but I don't think we can do better until and unless there is more detail.

[Joro Kolev]

For calculations and linear analysis see the user written -gtools- suite.

[Daniel Feenberg]

See http://www.nber.org/stata/efficient but generally the worst problems arise when there is insufficient physical memory. You can check the memory between procedures with the -memory- command, but some procedures need temporary storage. Watching the task manager (or equivalent in OSX or Linux) while a job is running will show you when the computer runs out of memory.

[Olivia Emma]

Thanks everyone

the data i am dealing with have 16,734,105 obs. I am trying to use the parallel computing

[Leonardo Guizzetti]

Unfortunately this is still vague, so no specific advice can be given. You have about 16 million records, and pretty much every operation takes more time with more data, regardless of which software you use. If there is something especially slow, please show us the code you are trying to optimize. There may (or may not) be faster ways to accomplish the same thing. In other cases, you simply have to wait.

Until then, there has been plenty of general advice offered here.