I generated a habitat distribution model using maxent software. I wanted to assess the accuracy of my model using TSS. I know that TSS is a threshold independent method, so using it means that I'm assessing a binary model (suitable, unsuitable). However I'm planning to present my model in 4 classes (highly suitable, moderately suitable, least suitable, unsuitable) using a boundary/class limit to separate the data into these classes in ArcGIS (Arcmap).
I was thinking that the last boundary - the one that separates the unsuitable class- is considered a threshold that I can use in conducting TSS? Since the other boundaries are between suitable classes - they differ in degree but still suitable-, so in other words the last boundary is separating suitable from unsuitable. Is this statistically correct?
question from:
https://stackoverflow.com/questions/65937256/tss-true-skill-statistic-threshold 与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…