SQLite database for storing an unlimited number of series: save with name, open, rename, delete, share and search.
How to work with the database:
The main screen has a bar at the top with the name of the open series or "New (unsaved) series". Below it are the Save and My Series buttons. Both actions are also in the menu.
Save: you enter a name and the series is saved along with the season length (m) and the selected method. If you save an already open series, it is updated. If the name is taken by another series, the application asks whether to overwrite it.
The "My Series" list:
Each row shows the name, number of values, last modified date, and the first values;
at the top there is a search by name;
Tapping on the series opens them, and holding them allows you to open, rename, share as text, or delete them (after confirmation); the newest are at the top;
The banner is down.
The last series is automatically opened upon startup. When first launched, the database already contains the sample series "Airline Passengers 1949–1952".
Help has a new "Saving Series" section in all 5 languages.
A library of example series has been added: real classical data sets plus synthetic series, each selected for a specific method. I first check the real data against their known averages.
14 example series have been added, each selected for a specific method. The button is called Examples and opens a list with the name, short description and number of values for each series. When selecting a series, the method and season length are set automatically.
Real classical data:
Airline tickets 1949–1952 (48 values) and the full series 1949–1960 (144), for Holt-Winters
Nile River discharge 1871–1970 (100), with level change around 1898
Canada lynx 1821–1934 (114), with a cycle of about 10 years, for ACF
Synthetic series. They are generated with a fixed seed, so they are always the same.
quarterly sales, m = 4 → Holt-Winters
daily visits, m = 7 → decomposition
linear trend + noise → regression
growing subscribers → Holt
stable search → simple exponential smoothing
temperature with noise → moving average
white noise → Lung-Box does not reject (p = 0.47)
random wandering → differentsirane
AR(1) with φ = 0.7 → CHAKF is interrupted after lag 1
MA(1) with θ = 0.8 → ACF breaks after lag 1