? Detrend（去除線性趨勢(shì)）— 從仿真，預(yù)測(cè)以及誤差數(shù)據(jù)中移除線性趨勢(shì)。此選項(xiàng)允許檢查實(shí)際數(shù)據(jù)的變化是否仍然被仿真或預(yù)測(cè)輸出所捕獲。例如，你觀察到你的仿真或預(yù)測(cè)數(shù)據(jù)偏離實(shí)際數(shù)據(jù)，但它們的變化似乎彼此一致。你可以通過消除仿真或預(yù)測(cè)數(shù)據(jù)的長(zhǎng)趨勢(shì)來檢查這個(gè)假設(shè)。當(dāng)選中detrend復(fù)選框時(shí)，將計(jì)算一個(gè)線性回歸模型。且將從仿真數(shù)據(jù)中減去線性回歸模型的輸出。Detrend選項(xiàng)適用于FIR，F(xiàn)2P和PAR模型的輸出。目前線性回歸模型不對(duì)每個(gè)數(shù)據(jù)段進(jìn)行計(jì)算，但對(duì)所有數(shù)據(jù)執(zhí)行計(jì)算。注意若當(dāng)你在應(yīng)用detrend之后仍然觀察到一些或所有數(shù)據(jù)段中有漂移，這可能意味著不同的數(shù)據(jù)段間的線性趨勢(shì)有不同。此外請(qǐng)確認(rèn)將空數(shù)據(jù)點(diǎn)標(biāo)記為錯(cuò)誤段，以避免不必要的壞擬合。
下面兩個(gè)圖顯示了去除線性趨勢(shì)功能的應(yīng)用。第一張圖顯示FIR，F(xiàn)2P和PAR模型的輸出偏離了實(shí)際數(shù)據(jù)。在這個(gè)例子中，輸入/輸出模型是斜坡模型。第二張圖顯示了原始數(shù)據(jù)和去除線性趨勢(shì)仿真數(shù)據(jù)之間的比較。在去除FIR和F2P模型的線性趨勢(shì)之后，仍然可以合理地捕獲原始數(shù)據(jù)的變化。
仿真窗口字段
Segments（數(shù)據(jù)段）：包含所有工作區(qū)數(shù)據(jù)段的下拉列表。
Input（輸入）：在預(yù)測(cè)圖上部子圖中的去趨勢(shì)輸入位號(hào)。
Output（輸出）：輸出及其預(yù)測(cè)值繪圖。
Sample Time（采樣時(shí)間）：當(dāng)前光標(biāo)位置的采樣時(shí)間。
Simulation/Error（仿真/誤差）：這些單選按鈕用于控制圖像是顯示仿真結(jié)果還是誤差結(jié)果。
Detrend（去除線性趨勢(shì)）：若選中此項(xiàng)，擬合輸出（仿真，預(yù)測(cè)，差分?jǐn)?shù)據(jù)，誤差或所有這些選項(xiàng)的組合）將被去除線性趨勢(shì)。
Difference（差分）：若選中此項(xiàng)，輸出和擬合輸出的時(shí)間序列會(huì)有差異。差分對(duì)于評(píng)估斜坡模型的擬合優(yōu)度非常有用。
圖例表
Show（顯示）：若選中此項(xiàng)，相應(yīng)的模型預(yù)測(cè)將顯示在預(yù)測(cè)圖上。
Color（顏色）：雙擊以修改預(yù)測(cè)圖線的顏色;此線條顏色也對(duì)應(yīng)于overlay和FIR fit窗口中的線條顏色。
原文：
**? Detrend **– Remove linear trend from simulation, prediction as well as the error data. This option allows one to check whether the variations of the actual data are still captured in the simulated or predicted output. For example, you observe that your simulation or prediction data drifts away from the actual data but their variations seem to agree with each other. You can check this hypothesis by detrending the simulation or prediction data. When the detrend checkbox is checked, a linear regression model will be calculated. The output of the the linear regression model will be subtracted from the simulation data. Detrend option applies to the output of the FIR, F2P, and PAR models. Currently, the linear regression model is not calculated for each data segment, but calculated for all data. Keep that in mind when you still observe drifts in some or all segments after applying detrend as it may suggest that the linear trends differ from one segment to the other. Also, ensure that you mark empty data points as bad segments to avoid unnecessary bad fits.
The two figures below show the application of the detrend feature. The top figure shows that the outputs of FIR, F2P, and PAR models drift away from the actual data. In this example, the input/output models are ramp models. The figure at the bottom shows the comparison between the original data and the detrended simulation data. Variations of the original data can still be reasonably captured after linear trends from FIR and F2P models are removed.
Simulation Window Fields
Segments: Drop down list with all the workspace segments.
**Input: **Input tag trended in upper subplot of prediction plot.
**Output: **Output and its predictions to be plotted.
Sample Time: Sample time of current cursor position.
Simulation/Error: These radio buttons control whether the plots show simulation results or error results.
Detrend:If checked, the fit output (either simulated, predicted, differenced data, error or the combination of all these options) will be detrended.
Difference: If checked, the time series of the output and the fit output are both differenced. Differencing is very useful for assessing the goodness of fit for ramp models.
Legend Table
Show: If checked, the corresponding model prediction is shown on the predictions plot.
Color:Double-click to modify the prediction plot line color; this line color also corresponds to the line colors in the overlay and FIR fit windows.
2016.11.23
