diff --git a/src/booster.jl b/src/booster.jl
index 648acca..5312019 100644
--- a/src/booster.jl
+++ b/src/booster.jl
@@ -275,20 +275,17 @@ function deserialize(::Type{Booster}, buf::AbstractVector{UInt8}, data=DMatrix[]
     deserialize!(b, buf)
 end
 
+
 # sadly this is type unstable because we might return a transpose
 """
     predict(b::Booster, data; margin=false, training=false, ntree_limit=0)
-
 Use the model `b` to run predictions on `data`.  This will return a `Vector{Float32}` which can be compared
 to training or test target data.
-
 If `ntree_limit > 0` only the first `ntree_limit` trees will be used in prediction.
-
 ## Examples
 ```julia
 (X, y) = (randn(100,3), randn(100))
 b = xgboost((X, y), 10)
-
 ŷ = predict(b, X)
 ```
 """
@@ -314,6 +311,58 @@ function predict(b::Booster, Xy::DMatrix;
 end
 predict(b::Booster, Xy; kw...) = predict(b, DMatrix(Xy); kw...)
 
+
+"""
+    predictbytype(b::Booster, data::DMatrix; type=0, training=false, ntree_limit=0)
+
+Use the model `b` to run predictions on `data`. 
+
+This version of predict gives access to contribution and interaction values.
+
+If `ntree_limit > 0` only the first `ntree_limit` trees will be used in prediction.
+
+The 'type' parameter conforms to prediction types specified in the XGBoost documentation.
+Options include:
+    0 => normal (default)
+    1 => output margin
+    2 => predict contribution
+    3 => predict approximate contribution
+    4 => predict feature interactions
+    5 => predict approximate feature interactions
+    6 => predict leaf training (see XGBoost documentation)
+
+The shape of returned data varies with 'type' option and certain objectives.
+
+## Examples
+```julia
+(X, y) = (randn(100,3), randn(100))
+b = xgboost((X, y), 10)
+
+ŷ = predict(b, X, type=2)
+```
+"""
+function predictbytype(b::Booster, Xy::DMatrix;
+                 type::Integer=0,  # 0-normal, 1-margin, 2-contrib, 3-est. contrib,4-interact,5-est. interact, 6-leaf
+                 training::Bool=false,
+                 ntree_lower_limit::Integer=0,
+                 ntree_limit::Integer=0,  # 0 corresponds to no limit
+                )
+    opts = Dict("type"=>type ,
+                "iteration_begin"=>ntree_lower_limit,
+                "iteration_end"=>ntree_limit,
+                "strict_shape"=>false,
+                "training"=>training,
+               ) |> JSON3.write
+    oshape = Ref{Ptr{Lib.bst_ulong}}()
+    odim = Ref{Lib.bst_ulong}()
+    o = Ref{Ptr{Cfloat}}()
+    xgbcall(XGBoosterPredictFromDMatrix, b.handle, Xy.handle, opts, oshape, odim, o)
+    dims = reverse(unsafe_wrap(Array, oshape[], odim[]))
+    o = unsafe_wrap(Array, o[], tuple(dims...))
+    length(dims) > 1 ? permutedims(o, reverse(1:ndims(o))) : o  # permutedims to handle ndims>=3
+end
+
+
 function evaliter(b::Booster, watch, n::Integer=1)
     o = Ref{Ptr{Int8}}()
     names = collect(Iterators.map(string, keys(watch)))