Update petrin.ipynb (#139)

fix typo describing the shape of inside good weights

docs/notebooks/tutorial/petrin.ipynb

@@ -657,7 +657,7 @@
    "source": [
     "We called the dataset \"CEX\", defined the number of observations in it, and also defined a lambda function for computing survey weights in a market. The `compute_weights` function has three arguments: the current market's ID $t$, the $J_t$ :class:`Products` inside the market, and the $I_t$ :class:`Agents` inside the market. In this case, we are assuming that each product and agent/consumer type are sampled with equal probability, so we simply return a matrix of ones of shape $I_t \\times (1 + J_t)$. This sets each $w_{dijt} = 1$.\n",
     "\n",
-    "By using $1 + J_t$ instead of $J_t$, we are specifying that the micro dataset contains observations of the outside option $j = 0$. If we instead specified a matrix of shape $I_t \\times (1 + J_t)$, this would be the same as setting the first column equal to all zeros, so that outside choices are not sampled from.\n",
+    "By using $1 + J_t$ instead of $J_t$, we are specifying that the micro dataset contains observations of the outside option $j = 0$. If we instead specified a matrix of shape $I_t \\times J_t$, this would be the same as setting the first column equal to all zeros, so that outside choices are not sampled from.\n",
     "\n",
     "We will be matching a few different statistics that were computed from this survey. For convenience, they are packaged in a data file with pyblp.\n"
    ]