Dyn4cast 11.11.26-alpha 2024-10-14
roxygen 2 misleading internal keyword demobilised to index all the help pages.
modelsummary’s datasummary_df insistence on tinytable object for some summary data addressed, The datasummary_df was demobilised.
Dyn4cast 11.11.24 2024-07-28
treatment_model introduced for propensity score matching and treatment effects
Dyn4cast 11.11.24-kappa 2024-07-02
PRESS suspended in MlMetrics
minor bugs discovered in the garret ranking function modules fixed.
Dyn4cast 11.11.24-kappa 2024-06-24
minor bugs fixed probably due to changes in dependencies following new R release.
Dyn4cast 11.11.24-beta 2023-09-29
Huge improvements and new additions to the package. A lot of editorial corrections observed in the codes have been carried out. Two new functions, Model_factors and garrett_ranking are successfully tested and now operational. The two functions provides for gaining deeper insights into the meaning behind Likert-type variables collected from respondents. Garrett ranking provides the ranks of the observations of the variables based on the level of seriousness attached to it by the respondents. On the other hand, Model factors determines and retrieve the latent factors inherent in such data which now becomes continuous data. The factors or data frame retrieved from the variables can be used in other analysis like regression and machine learning.
The two functions are path of factor analysis, essentially, exploratory factor analysis (EFA), used to unravel the underlying structure of the observed variables. The analysis also helps to reduce the complex structure by determining a smaller number of latent factors that sufficiently represent the variation in the observed variables. With EFA, no prior knowledge or hypothesis about the number or nature of the factors is assumed. These are great tools to help tell the story behind your data. The data used for Model_factors is prepared using fa.parallel and fa functions in the psych package.
Dyn4cast 11.11.24-alpha 2023-09-25
The Data argument for DynamicForecast function has been modified. Rather than the previous argument were a data frame, containing the two required vectors, was supplied, now data argument is in three vectors date, series, x. The x argument is new but optional and would now allow additional data apart from the two mandatory ones to be added to the model. The origin argument which was previously embedded in ... has now being defaulted to 1970-01-01. Two objects have been added to the results, i.e., table of estimates and the plots of the fitted values. The function, which was the first in the package has now be promoted to stable.
Dyn4cast 11.11.23
The new version is coming with a huge improvement. The number of functional forms of the linear regression has increased from seven to 14 based on the Taylor Series expansion of the Box-Cox transformations. In addition, two new plots are added to the suite i.e. fitted and effects plots. Also, Tables of marginal effects are now produced for fast inference from the estimated coefficients. This new version also comes with 47 metrics for assessing model performance and carry out model diagnostics. This new version also comes with the capability of making prediction about each of the functional forms.
There are two new functions coming along with this new version. corplot for plotting the correlation matrix of the variables included in the model. estimate_plot is for plotting the estimated coefficients in their order of significance in the estimated model. In addition to a new plot, visual means it has become lots easier to understand the nature of the distribution of the variables, digest and make inference about the model.
Even with this huge improvements, the principle of one line of code is maintained. It is easy and can be implemented with less headaches once the data has been cleaned. The structure is Linearsystems(y, x, mod, limit, Test). The argument Test is optional, if omitted, the prediction is made with NULL data.
As always, comments, questions and suggestions are welcome.
Dyn4cast 11.11.10-kappa 2023-04-02
quicksummary now returns a data frame rather a kable Table.
The internal argument digits in Percent is removed due to observed exaggeration of the final rates
Final formatting of data frames in the Linearsystems package was removed. Now, the estimates can display in the console.
Dyn4cast 11.11.10-beta 2023-03-12
A new function data_transform added and is working perfectly. The addition of this function only makes the package more versatile for Machine Learning operations and estimations.
Dyn4cast 11.11.10-alpha 2022-08-20
A new function formattedcut added and is working perfectly. The addition of this function is providing more clarity to the package. Quite a number of editorial fixes were undertaken and the codes are working superbly.
Dyn4cast 11.11.1.9000 2022-04-25
Mallow’s Cp successfully added and documentation improved sufficiently. The codes are working better now and the check results are encouraging.