I will try to explain the differences and when to use them. On Tableau, you can download the Superstore sample dataset and put sales on Columns and Profit on the Y-axis to follow along.
If we choose the circle under the marks card we get a scatter plot.
Here this would try to see the relationship between sales and profit. The linear graph can sometimes be unrealistic, for example, the y-intercept is negative.
Linear trend line. The name derives from its shape, an exponential or rapid increase. A logarithmic equation is similar to the inverse of the exponential function. Here Ln represents the natural logarithm or log base e.
The graph tends to rise very quickly and then trails off slowly rising as the x value increases. The power trend line does not have a constant and is dependant on the x-axis value and a constant determined by tableau. It may look similar to the linear function in the graph, but it is created from a completely different function.
The R squared value is always between 0 and 1 and represents the explained variation divided by the total variation. Going through the iterations of the different trend lines using the superstore data between profit and trend line type I found the power trend line to have the highest R squared value. Of course, I would not recommend using the trend line that has the highest R-value all the time, as visualisation is the most important thing trend line type consider.
If it looks nice and matches up with a lot of the points and is easy to comprehend then use it. Also, use some intuition. If the graph looks like it slopes upwards in a curve fashion then a linear trend line is not the best to use. However, a linear trend line does have the advantage to be able to clearly see whether the trend is upwards or downwards sloping, and what the magnitude is easier than other lines. P-Value There is P or probability value assigned to the trend line.
The lower this value the better.
Excel trendline types
Without going into stats too much, Tableau is computing how well your trend line fits the model based on an F statistic with a hypothesis that an intercept only model is accurate. If P is low we reject this and fail to reject or accept, but we can never truly accept a hypothesis the alternate option that this trend trend line trend line type accurately describes the data.
A low p value may be a bad thing as the line may not be the best model to follow. To put it another way, trend line type is a 1-p chance that the coefficient of the trend line is not 0.
Updated Nov 16, What Is a Trendline? Trendlines are easily recognizable lines that traders draw on charts to connect a series of prices together or show some data's best fit. The resulting line is then used to give the trader a good idea of the direction in which an investment's value might move. A trendline is a line drawn over pivot highs or under pivot lows to show the prevailing direction of price. Trendlines are a visual representation of support and resistance in any time frame.