We saw that buckets at District and Block level seem to be positively correlated with Reading and Math baseline levels. At a DISE level, we do not see a strong correlation of buckets with Reading or Math features(BL and Growth). Thus, we are looking at correlation between Reading and Math baselines to if they are positively correlated
When Comparing Baselines of Math and Reading, we see that:
Should we consider doing clustering at a DISE level for Maths and Reading separately?
We see that all these blocks belong to the medium bucket. We pick one such block to compare BL and Growth features of DISE to the block
## # A tibble: 1 x 2
## bucket_block `n()`
## <chr> <int>
## 1 medium 113
For the Block BAGPAT(a medium block with 80% DISE schools in low bucket), we are looking at how the features of the block compare with the features of the district.
From the scatter plot, we see that because of some extreme points, the features at a block level are closer to the average of the dise features.
If we compare the block level reading baselines and growth, we see that they are closer to the means of the dise-level features than their medians
| PrathamBlockName | validation_feature_reading_bl_1_block | growth_rd | mean_dise_bl_rd | median_dise_bl_rd | mean_dise_growth_rd | median_dise_growth_rd |
|---|---|---|---|---|---|---|
| BAGPAT | 0.0769231 | 0.315981 | 0.0616125 | 0 | 0.3286321 | 0.2891705 |