TAYSTUDIO operation notes · tool guides · tech/career · useful posts
![[AWS Infra] Building & Configuring a NAT Instance from an AMI to Cut NAT Gateway Costs (Part 1)](assets/posts/nat-instance-build-1/cover.webp)
A NAT Gateway costs $32+ a month, so I connected a private subnet to the internet with an EC2 NAT instance instead — Part 1 uses the AWS community NAT AMI.
![Implementing the t-test (t test) [One, Two, Paired Samples t-test: Python feat. SciPy & Statsmodels]](assets/posts/t-test/cover.webp)
Turning t-test theory into SciPy code — ttest_1samp, ttest_ind with the equal_var toggle, and ttest_rel — and verifying the hand-computed t and p match SciPy.
As sample size grows, the sample mean converges to a normal distribution for any population — verified across five distributions, plus why the CLT matters.
How to tell if a difference between two group means is real or just random variability, plus the t-test's assumptions and worked calculations for each type.
Get started with Python FastAPI — why async and the GIL matter, pip install uvicorn, ASGI vs WSGI, run the server two ways in PyCharm, plus built-in Swagger UI.
When overwriting PATH on Linux or macOS breaks commands like ls, vi, and vim with 'command not found' — check echo $PATH and restore base paths via export.
Why Windows CMD's cd can't switch drives and you must type 'D:' to move from C: to D — Windows gives each drive its own root, unlike Linux's single root.
Install Python on Windows and add the install and Scripts (pip) paths to your PATH so 'python' and 'pip' work in cmd — plus user vs system variables explained.
Why confidence intervals exist, how they connect to hypothesis testing, and the truth behind the myth that there's a 95% chance the mean is in the interval.
A 2019 undergrad thesis using a neural network (ANN) to predict PM2.5 levels 6-12h ahead — 94,225 observations, AIC/VIF selection, and a time-lag pitfall found.