--menu
Help
Exit
--opciones.1
Exploratory analysis
Transformations
Tutorial
--opciones.2
Series manager
Brief explanation of an Stationary Time Series
Exploratory analysis
Non-constant variance
Seasonal component
Non-constant mean
Model detection
Back to previous menu
--opciones.3
Choose series
Delete series
Back to main menu
--opciones.4
Series manager
ACF/PACF
Insert model
Model estimation
Back to previous menu
--opciones.5
Model manager
Estimation
Fix part to zero
Modify model
Models validity
Back to previous menu
--opciones.6
Choose model
Delete model
Back to main menu
--opciones.7
Model manager
Residual analysis
Compare theoric-sample ACF/PACF
AR(infinite)/MA(infinite)
Delete model
Prediction capacity
Back to previous menu
--opciones.8
Model manager
Estability
Prediction capacity
Choose best model
Atypical treatment
Long-term predictions
Back to previous menu
--opciones.9
Model manager
Atypical treatment
Back to previous menu
--opciones.10
Model manager
Long-term predictions
Back to previous menu
--previo.1
PREDICTIONS
TRANSFORMATIONS
MODEL DETECTION
MODEL ESTIMATION
MODEL VALIDITY 
PREDICTION CAPACITY
ATYPICAL TREATMENT
LONG-TERM PREDICTIONS
PREVIOUS
--previo.2
Predictions
Transformations
Model Detection
Model Estimation
Model Validity
Prediction Capacity
Atypical Treatment
Long-term Predictions
Previous
--previo.3
The objective of the program is to get predictions of the series
which starts in
and with a period
. Firstly, it is  recomended to make an exploratory analysis to get a previous idea of the series behaviours and properties. Then, the next step is 'Transformations'.
To can obtain models to get good predictions, the series has to be stationary (see the vignette 'Stacionarity' or select the option 2). For this, if it necessary, you have to make certain transformations. Firstly it has to check the uniform of the variance, if it is not uniform, in most cases, it is solved applying logarithm.
Once we have an stationary series we can identify its possible models. To can do it, you have to look its ACF/PACF plot and, then, introduce the found models
Once we have some models, we have to estimate them and, in case it has no significative coefficients, depends on the case, you have to fix these coefficients or modify the models until they have significative all their coefficients.
When we have significative models, the last step is the validation where we may rule out the invalid models.
With the obtained valid models you can obtain predictions but it can be some differences between the quality (Reality aproximation) in these predictions. Also, these models could not be steady (The same model fitted without the lastest observations has coefficients very different). For this, it is recomended analyze the stability and the prediction capacity of the different valid models to then choose the best model and proceed to the 'Atypical Analysis' or to get 'Long-term Predictions'. The advisable number of reservations is one period of the series. 
The atypic treatment allows to obtain the series that would exist if the unexpected events, which have caused changes in the series, hadn't happened. In case to use the atypical treatment, with the obtained series, the program readdress to the menu 'Transformations' to restart (The linealized series mantains the transformation applied to obtain constant variance). For the atypical detection be optimal, it is necessary that the active model was the best model created.
Once you have a good model, you can obtain long-term predictions. Remember that the confident intervals are at 95% of confidence.
--ayuda.1
'Initial' help menu
1: The 'Exploratory analysis' option makes a data analysis showing numeric and graphical results.
2: The 'Transformations' option readdress to the following menu where you can make transformations to the series for making it an stationary one.
3: The 'Tutorial' option opens a .pdf file which contains a brief tutorial of TSTutorial.
--ayuda.2
'Series manager' help menu
1: The 'Choose series' option let choose, from a list which contains all the series, the series which you want to work.
2: The 'Delete series' option let delete one series from the list of series which contains all the created series.
--ayuda.3
'Transformacions' help menu
1: The 'Series manager' option lets to manage the series choosing with which work or delete.
2: The 'Brief explanation of an Stationary Time Series' option opens a .pdf file which shows with examples how is an stationary series and the differents non-stationary series that it could be founded.
3: The 'Exploratory analysis' option makes a data analysis showing numeric and graphical results.
4: The 'Non-constant variance' option shows, through some analysis, whether the series has constant variance and, if it is necessary, you have to indicate what transformation you want to make.
5: The 'Stationality' option shows whether the series can have stationary component for, then, make it an stacional differenciation, in case it was necessary.
6: The 'Non-constant mean' option shows whether the series has non-constant mean and, if it is necessary, makes a regular differenciation.
7: The 'Model detection' readdress to the next menu where you can identify and introduce the possible models of the series.
--ayuda.4
'Model detection' help menu
1: The 'Series manager' option lets to manage the series choosing with which work or delete.
2: The ACF/PACF option shows the ACF/PACF of the active series to let you to deduce its possible models.
3: The 'Insert model' lets introduce one possible model of the active series.
4: The 'Model estimation' readdress to the next menu where you can estimate and modify the created models.
--ayuda.5
'Model manager' help menu
1: The 'Choose model' option let choose, from a list which contains all the models, the model which you want to work.
2: The 'Delete model' option let delete one model from the list of models which contains all the created models.
--ayuda.6
'Model estimation' help menu
1: The 'Model manager' option lets to manage the models choosing with which work or delete.
2: The 'Estimation' option shows the active model estimation, suggest whether it has to work with constant and whether it contains non-significative coefficients.
3: The 'Fix part to zero' option lets to modify one coefficient on each time you select the option. Then, it contrats the fixed model with the previous one to confirm the action.
4: The 'Modify model' option modify the active model letting to choose what componoents you want to modify (p, q, P o Q).
5: The 'Models validity' readdress to the next menu where, through some analysis, you can determine whether the active model is valid to make predictions.
--ayuda.7
'Model validity' help menu
1: The 'Model manager' option lets to manage the models choosing with which work or delete.
2: The 'Residual analysis' option shows through some plots whether the analized model has good residuals and, in this case, if the model is valid.
3: The 'Compare theoric-sample ACF/PACF' option compare the ACF/PACF's active model with the theorics to see if their behaviour are similar.
4: The 'AR(infinite)/MA(infinite)' option shows a comparative of the AR(infinite)/MA(infinite) from all models to can decide the existance of equivalent models.
5: The 'Delete model' option let delete the active model.
6: The 'Prediction capacity' option readdress to the next menu where you can calculate the stability and the prediction capacity of the models to decide which is the best model.
--ayuda.8
'Prediction capacity' help menu
1: The 'Model manager' option lets to manage the models choosing with which work or delete.
2: The 'Estability' option calculate if the active model is stable.
3: The 'Prediction capacity' option calculate the model prediction capacity calculating its Mean Squared Error.
4: The 'Choose best model' option assign the choosed model as the best model.
5: The 'Atypical treatment' option readdres to the next menu where you can apply the atypical treatment to linealize the series.
6: The 'Long-term predictions' option readdress to the next menu in which you can obtain long-term predictions.
--ayuda.9
'Atypical treatment' help menu
1: The 'Model manager' option lets to manage the models choosing with which work or delete.
2: The 'Atypical treatment' option indicates what criterions can be the best to use and you can linealize the series with the desired criterion.
--ayuda.10
'Long-term predictions' help menu
1: The 'Model manager' option lets to manage the models choosing with which work or delete.
2: The 'Long-term predictions' makes long-term predictions.
--ini.1
It shows the series numeric descriptive; the series plot, its histogram, the qqnorm plot and its ACF; and the descompose plot which shows the series without its seasonal component and trend, the series whithout its seasonal component, its trend and its remmember plot (reorderet ACF).
EXPLORATORY ANALYSIS OF THE SERIES:
Exploratory analysis of the series:
General numeric description:
The mean of the series is
and its standard desviation is
Its median value is
, its minimum value is
and its maximum value is
--gesser.1
Series list
Variance
Indicate the list number which contains the series with which you want to work.
The number that you have choosen it doesn't exist in the list.
--gesser.2
Indicate the list number which contains the series with which you want to delete.
The number that you have choosen it doesn't exist in the list.
(Remmember that you can't delete the original one)
The deleted series was the active series, for this reason you have to choose the series with which you want to work.
You have to have at least 2 series to can choose one to delete.
--trans.4
It shows a plot which contains the boxplots for every period, if you see that the initials are enought different from the final ones, its very possible that the series doesn't have constant variance. In case to transform the series, the program will make atomatically an exploratory analysis.
Nonconstan variance series:
SUGGESTION: It is suggested to transform the data.
SUGGESTION: It seems no necessary transform the data.
Write 0 if you want to make a logarithm transformation or the lambda value of the transformation. If you don't want to make a transformation write 1.
The transformation done is logarithmic.
It hasn't made a transformation.
It has made a transformation with lambda
Do you consider that the series is stationary?
The series
it is considered stationary.
it isn't considered stationary.
The active series has been transformed yet. Please, select a not transformed series.
--trans.5
As well as helping with the 'Exploratory analysis' monthplot, it also creates a new plot which contain one period with all the observations separated by position of the period that it take up. With it, you can see if it contains different heitghts is the key of the possible existance of stationary component. In case to want to differenciate the series, the seasonal order is the period of the series.
Series seasonal:
Do you want to differenciate?
What is the seasonal order of the series?
It is made a seasonal differenciation with order
to the series.
Do you consider that the series is stationary?
It is decided not make a seasonal differenciation to the series.
The series has been seasonal differenced before. For R specifications, R can't work with more than one seasonal differenciation.
You have written a negative number or is not integer.
--trans.6
You can have help using the monthplot from 'Exploratory analysis'. If the trend doesn't seems like a horizontal straight line is possible that you have to make a regular differenciation.
Non-constant mean series:
Do you want to differenciate?
It is made a regular differenciation to the series.
It is decided not to make a regular differenciation to the series.
Do you consider that the series is stacionary?
--iden.2
It generates a plot which contains the ACF and PACF of the series.
ACF/PACF of the series:
--iden.3
Wirte p
Wirte q
Wirte P
Wirte Q
Insertion the model of the series:
The model that you have introduced doesn't achieve the stacionality properties, try with another model.
It is created the model
You have written a negative number or it is not an integer.
--gesmod.1
Model List
Indicate the list number which contains the model with which you want to work.
The number that you have choosen isn't in the list.
You have to create at least one model to choose.
--gesmod.2
Indicate the list number which contains the model with which you want to work.
The number that you have choosen isn't in the list.
The delete model was the active model, for this reason you have to choose the model with which you want to work.
You need at least 1 model to can choose one to delete.
--estim.2
Model estimation of
The model constant isn't significative and this is the model estimation without constant:
SUGGESTION: The model AIC wihtout constant is less than with constant, for this reason, the model is better.
SUGGESTION: The model AIC wihtout constant is bigger than with constant, for this reason, the model is worse.
Do you want to work wihtout constant?
It is choosen to work without constant.
It is choosen to work with constant.
SUGGESTION: The
model has at least one final coefficient not significative. In case you want to fix it, is better to modify the model substracting one of the component which has this coefficient.
model has at least one final and middle coefficients not significative. In case you want to fix one of the final coefficients, is better to modify the model substracting one of the component which has this coefficient. On the contrary, you may fix the intermediate nonsignificative coefficients.
model contains middle coefficients not significative. It is recommended to fix it.
You have to have at least 1 model to can estimate it.
new
--estim.3
Sometimes the models contains nonsignificative coefficients (|Coef/s.e.|<1.96) that it has to fix to zero to make the model better.
Really do you want to fix one model coefficient?
Fix part of the model
Active model
SUGGESTION: It is recomended fixing the coefficient less significative (the closest to zero).
Indicate the position number that ocupates the coefficient you want fix (left->right).
You have written an incorrect position.
The choosen coefficient is already fixed to zero.
It is decided to fix the coefficient that ocupate the position
Modificate model
The modificated model AIC is less than the active, for this reason the model has improved.
The modificated model AIC is bigger than the active, for this reason the model has deteriorated.
Do you want to work with the modified model?
It has choosen to work with the modified model.
It has choosen not to work with the modified model.
SUGGESTION: The
model has at least one final coefficient nonsignifficative. In case you want to fix it, is better to modify the model substracting one to the component which have this coefficient.
model has at least one final and middle coefficients not significative. In case you want to fix one of the final coefficients, is better to modify the model substracting one of the component which has this coefficient. On the contrary, you may fix the intermediate nonsignificative coefficients.
model contains middle coefficients not significative. It is recommended to fix it.
You cannot fix more coefficients because the model only contains one.
You need at least 1 model to can estimate it.
new
--estim.4
Do you want to modify the model?
Modify the model
What parameter do you want to modify?
What parameter do you want to modify?
It is decided to modify the parameter
The actual value of p is
. Write its new value.
The new value of p is
The actual value of q is
The new value of q is
The actual value of P is
The new value of P is
The actual value of Q is
The new value of Q is
The model that you have introduced doesn't achieve the stacionality properties, try with another model.
You need at least 1 created model to can modify it.
You have written a negative number or it is not integer.
--valid.2
Generates 6 plots: the model residuals, the data aproximation to the Normal straight line, the residual histogram, the residual ACF/PACF, the squared residual ACF/PACF and the Tsdiag
SUGGESTION: The analysis indicates that the model is valid.
SUGGESTION: The analysis indicates that the model is not valid.
Do you consider that the model is valid?
The model
is considered valid.
is not considered valid.
You need at least 1 created model to can analyze the residuals.
Residual analysis of
--valid.3
Comparation of the theoric ACF/PACF with the sample
Generates 2 plots, the first contains the ACF and the second the PACF.
You need at least 1 created model to can comparate the theoric ACF/PACF with the sample.
--valid.4
Shows in the console the invertible coefficients and creates a plot that contains the model expressions like AR and MA inifinte.
Model AR/MA inifintes
Invertibility coefficients
SUGGESTION: The model is invertible.
SUGGESTION: The model is not invertible, for this reason, it is advisable to work with another model because its predictions will not be very specified.
You need at least 1 created model to can see the AR(infinite)/MA(infinite).
Your model is an AR model. By definition, your model is invertible.
--valid.5
Do you want to delete the active model?
--cap.2
How many observations do you want to reserve?
You cannot reserve more observations than the length of the series or less than one and a non-integer number.
Model estability of
It has been choosen reserve
observations.
A stable model is when the same model without the reserved observations mantains similar values of the original model coefficients. In  case to be a quite a lot of differents it can be an indication of the atypical presence in the last part of the series.
SUGGESTION: It seems that the model is a quite a lot of unstable because many coefficients have changed considerably in its values. You should choose another model to analyze its prediction capacity because its MSE will be very high. Another option is to make an 'Atypical treatment' and then draw conclusions.
SUGGESTION: It seems that the model is a little bit unstable, it could be for the existance of atypical values in the lastest observations from the series.
The linealize criterion may be smaller.
SUGGESTION: All the tests indicates that the model is stable.
Do you consider that the model is stable?
It is considered that the model is stable
It is not considered that the model is stable
You need at least one created model to can study its stability.
--cap.3
Generates a list shown in the console and a plot that contains the obtained reserved predictions with its confident intervals and the real data of it observations.
Model predict capacity of
How many observations do you want to reserve?
You cannot reserve more observations that the length of the series or less than one and a non-integer number.
It has been choosen reserve
observations.
Obtained predictions with the original data
You hav
You need at least one created model to can check its prediction capacity.
--cap.4
To select the best model, it useful use the plot which shows the model collection to can compare it.
Best model selecction
Do you want to continue with the best model selection?
SUGGESTION: It useful analyze the prediction capacity of all the models to have an optimal comparation between it.
SUGGESTION: The model with best AIC is
Write the position taken up in the model list that you want to fix as best model.
You have written an incorrect position.
It have been choosen as best model
Do you want that the best model was the active model?
SUGGESTION: The model with best AIC and MSE is
SUGGESTION: The model with best MSE is
The unic created model that you have is already selected as best model.
You have only one created model. Do you want to choose it as best model?
You need at least one created model to can choose it as best model.
--atip.2
Creates two plots to make a comparative making different treatments, one have the atypicals with LS and the other one wihtout LS. You have to compare between the plots and between the results in the plots.
This procedoure could take some time....
Do you think that you have to make an atypical treatment to the series?
Atypical treatment of the model
Write the treatment criterion:
Do you want that the treatment works with LS?
The treatment criterion is
with
without
LS.
Detectected atypical list
With the choosen criterion, it hasn't found any atypical.
The linealized series variance is
You cannot make the Atypical treatment with a model already linealized.
You have to be al least one created model to can apply it the Atypical treatment.
The criterion has to be bigger than 0.1.
--prev.2
Generates a list shown in the console and a plot that contains the obtained predictions with its confident intervals and the previous series data.
Lon-term predictions of the model
How many predictions you want to obtain?
You need at least one created model to can obtain its long-term predictions.
Long-term predictions
The number has to be a positive integer.
--salir
Are you sure that you want to exit?
Session Data Summary
--plot
(Plot,Histogram,Qqnorm,Acf)
(Plot descompose)
(Boxplot)
(Monthplot)
(ACF/PACF plot)
(Residual plot)
(Residual QQnorm plot)
(Residual histogram)
(Residual ACF/PACF plot)
(Squared residual ACF/PACF plot)
(Tsdiag plot)
(ACF sample vs theoric plot)
(PACF sample vs theoric plot)
(Capacity of Prevision plot)
(Series minus linealized series plot)
(Superimpose the series with the linealized series plot)
(Long-term predictions plot)
(List of series plot)
(List of models plot)
(AR/MA infinite)
(Plot Mean vs Standard deviation)
(Plot Box-Cox transformation)
--drawser
SERIES COLLECTION
Series
Variance
Stat.
Lineal
no
and
The active series is 
--drawmod
MODEL COLLECTION
Arima model
Int.
AIC
Valid
Sta.
MSE
Lineal
no
Crit
The active model is
The best model is
MODEL COLLECTION
It doesn't exist any created model.
--writelistser
(Stationary)
(Linealized)
--writemod
with constant.
--writelistmod
: Arima model
(Validate)
(Linealized)
(Best model)
--writecoef
Model coefficients of:
Significative?
no
nearly
yes
--drawarma
with constant.
--atipics
SUGGESTION: It is not necessary the atypical treatment
SUGGESTION: With LS it is recommended to work with
SUGGESTION: Without LS it is recommended to work with
the criterions
the criterion
--drawatip
Atypical analysis with LS
Atypical analysis without LS
--previsiones
FITTING TEMPORAL SERIES
You have your report already in a .pdf file in your working directory.
--enterComment
Do you want to write a comment?
Following, write your comment.
COMMENT
--latexCntrl
You have choosen that TStutorial make a report, but you don't have installed any .tex compiler. For this reason, you cannot get it. Automatically, the function starts to work but without making the report.
--end