For some cases, by example when you works on webservices or an ETL, you can be a little lost between all the variables you use, specifically arrays.
I developed a derivative of the Hungarian notation to prefix my variables and find myself there. This notation can be applied in php, but also to any other languages, especially non-typed.
There is the prefixing I use with PHP variables in that specific cases, depending on their type and design :

Scalar types prefix

• $is, $b : boolean
• $s : string
• $i : int (unsigned)
• $n : natural (signed int)
• $f : float
• $fd : float round to 1 decimals
• $fc : float round to 2 decimals
• $fm : float round to 3 decimals
• $m : mixed.

Non-scalar prefix types

• $o : object
• $r : resource (PHP-specific)

Array, can be, by design, a collection of elements, or a single element. For me an “element” is an object-like data, stored as an associative array. While a collection if an array of elements.
So I can distinguish 3 different types for arrays :

• $e : (array) Element (or Entry, or Entity) ; associative array containing data about an element.
• $c : (array) Incremented Collection ; an array with auto-incremented keys containing elements. Means that the array keys don’t identify an element, event if the order to process it can be important.
• $a : (array) Associative collection ; an array with associative & manually-assigned keys (like “hashes” in Perl ) containing elements.

For collections, I usually also put the variable in plural (or simply suffix with an s). I suggest to combine collection prefixes with the other prefix showing the values each entry contains :

• $ca : incremented-collection of Associative collection
• $cm : incremented-collection of mixed elements
• $ae : associative-collection of array elements
• $as : associative-collection of string elements
• $am : associative-collection of mixed elements (each entry can contain different things)
• $cbo : increment-collection of boolean OR objects. So we have to test if the entry is an object before processing it.

Semantic types :

• $u : unsafe type or value, need to be checked or sanitized for security reason.
• Can be combined : ie $us for unsafe string, with possible code injection.
• $zo : string of encoded object
• $zj : string of encoded json
• $zh : string of encoded html entities, ie: ‘"’
• $ts : timestamp (signed int)
• $d : date, format ‘Y-m-d’ (string)
• $dt : datetime, format ‘Y-m-d H:i:s’ (string)

I personnally try to avoid to prefix private members / functions with underscores.