When I compute matrix inverse on Python, it gives me slightly different results from Excel (difference appears after the third or forth number, however it highly affects my work). I know that for a non-square matrix there is no unique solution, hence, do you know a Python function that computes a matrix inverse using the same logic that Excel applies?
Excel is known to have some anomalies when it comes to the rounding of numbers. You may be best off trusting the Python approach more, depending on how you are approaching this, and the precision required for the values.
See e.g.: https://answers.microsoft.com/en-us/msoffice/forum/msoffice_excel-mso_other-mso_2010/matrix-inversion-producing-more-errors/c6bc6331-bbed-40f4-b6b0-099f1a288c1c
https://stackoverflow.com/questions/31256252/why-does-numpy-linalg-solve-offer-more-precise-matrix-inversions-than-numpy-li
https://www.r-bloggers.com/dont-invert-that-matrix-why-and-how/
https://support.office.com/en-us/article/minverse-function-11f55086-adde-4c9f-8eb9-59da2d72efc6
Hello Yasmine,
Microsoft excel is not designed as a scientific matrix calculation program.
The best way is to the inversion in Python or Mathlab and transfer the numbers to Excel after the calculation.
If you still wanted to use Excel than you could try this methods :
Define the matrix as double precision in a user macro and run the calculation from there. Another method could be : transfer the matrix to multiplied numbers by 10e8 and use the integer parts. (you have to define this numbers in the user macro as Long integers.
Excel is meant to present data and do somewhat simple calculations. It can do a lot of simple calculations quickly, but complex calculations will have a greater error from round off and propagation, The built in functions are meant for speed not precision, so I would not try to replicate the results.
Hi All,
Many thanks for your contribution.
I realized that the matrix I'm trying to get its inverse is of really high values and thus, I guess this was the problem.
To add to previous answers, there is a fundamental reason: algorithms used may be different in Python and Excel, thus leading to different roundoff errors and results.
Furthermore, roundoff itself (rounding or truncating) may differ on different processors, thus leading to different results even within a same application.
Dear Yasmine Megahed,
I'm stuck ini this problem too, the inverse of matrix that i'm trying to get is really big, (matrix 30x30). Can you please tell me what should I do with this case?:( Anyway, im using numpy library and it gives different solution from excel. Which one i can trust?
Dear Mentari Adiza Putri Nasution
Use Numpy and Python. As several answers above have told you Excel cannot be trusted with this calculation for several reasons. If in doubt, check manually the calculation of the inverse, it's only 30x30 which not that big. I've worked on 100x100 or bigger inverse matrix calculations, although using a different programming language.
Jan Krabbenbos Okk i got it, thank you for your help sir🙏🏻
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Programming Languages
- Python