26 Jul 2022Tuesday

Reorganized my Harvard Wiki page a little bit:
- Questions for Rubin Obs. → page to list questions about inconsistencies in data from AuxTel
- Data Clipping → page to keep track of how I am processing the data
- Stubbs Notes → page to keep track of Stubbs' comments/additions to my page
- Task Lists → page to log what I need to do, instead of putting it into the lab notebook
- Lab Notebook is now organized by week
Completed the "Basic Instruction" for machine shop training and sent out the certificate–waiting to hear more info about the RED certification class potentially next Wednesday.
Worked on Data_Clipping Jupyter Notebook to clean up code and commenting

Finding Covariances for Coefficients of Polynomial and Linear Models

In Data_Clipping.ipynb
worked on finding the covariances of the coefficients of the polynomial and linear model
- For the linear model, I used np.polyfit(airmasses, eq_widths, 1, cov=True). Returns coeffs in order of highest power and covariance matrix where diagonal values are the covariances of each of the coeffs
- created func to extract diagonal values and store them
- For polynomial func ax^(1/2) + bx + c, used curve_fit from scipy, also extract diagonal values from covariance matrix to each coeff
Created a table that gives the polynomial model and linear model for each molecule on each night and each star

Worked on the following sanity checks:
- Oxygen → independent of star and night
- Hydrogen → independent of night, plotted against H//alpha and H//beta
- Water → independent of star
When I looked at independence of molecule stars, I plot all eq_widths of stars against airmass on a given night
When I looked at independence of molecule and night, I plot all eq_widths of molecule for each star
Mimicked code from Data_Clipping.ipynb in new nb Mol_Independence.ipynb
Created a function that takes in the molecule in question ('$O_2$', "$H_2 O$', etc.) and the needed independence factor (ie. "date", "star") and outputs:
1. any errors (No HDU table, neg eq_width)
2. table of linear model and covariance of coefficients for each star or date
3. scatterplots for airmass and eq_width

Here are the results of running Jupyter nb Mol_Independence.ipynb:

Worked on adding the error bars to the data using mpl. I remembered midway trying to figure out error bars that in previous meeting Chris mentioned that the error reported on the equivalent widths were not provided (and something we should ask about).
Still managed to plot the standard error of all the eq_widths, the error bars were incredibly large when taking std/ sqrt(n). Talked with Eske, and he recommended that I only plot std, which made error bars look much nicer
Worked on applying error shading bands on the data, and it looks a bit clunky, will need to figure out how to plot the error bar shading bands in numerical order (may require ordering the list of airmasses, but then I have to also order the list of eq_widths)
Also, weirdly the molecule data for O2 became orange. Currently working in the Mol_independence.ipynb to add error shading bars

Next steps to consider: fix orange O2, fix error shading bands, find R^2 values for the linear fits, and do preliminary statistical analysis for independence (using the linear fits). Also can start extension project

Re-did the linear fits, this time using scipy.stats.linregress instead of np.poly_fit
scipy easily gives the r^2 value, but np.poly_fit gives the covariance of each variable coefficient. Overall, I prefer using scipy.stats.linregress, easier to index into
For finding the independence of specific air molecules against each star or night, I added the linear fit model and I wrote down the model and r^2 value on each plot.
It looks like the linear fits are pretty weak for the data because the spreads are so high, and the r^2 values are pretty close to zero, so I'm not sure if I can use the linear model to determine independence between the air molecules eq_widths and airmass
Example data plots, note very low r^2 values